1. Cell Biology
  2. Neuroscience
Download icon

Glycolytic reliance promotes anabolism in photoreceptors

  1. Yashodhan Chinchore
  2. Tedi Begaj
  3. David Wu
  4. Eugene Drokhlyansky
  5. Constance L Cepko  Is a corresponding author
  1. Harvard Medical School, United States
  2. Howard Hughes Medical Institute, Harvard Medical School, United States
Research Article
  • Cited 29
  • Views 2,010
  • Annotations
Cite this article as: eLife 2017;6:e25946 doi: 10.7554/eLife.25946

Abstract

Vertebrate photoreceptors are among the most metabolically active cells, exhibiting a high rate of ATP consumption. This is coupled with a high anabolic demand, necessitated by the diurnal turnover of a specialized membrane-rich organelle, the outer segment, which is the primary site of phototransduction. How photoreceptors balance their catabolic and anabolic demands is poorly understood. Here, we show that rod photoreceptors in mice rely on glycolysis for their outer segment biogenesis. Genetic perturbations targeting allostery or key regulatory nodes in the glycolytic pathway impacted the size of the outer segments. Fibroblast growth factor signaling was found to regulate glycolysis, with antagonism of this pathway resulting in anabolic deficits. These data demonstrate the cell autonomous role of the glycolytic pathway in outer segment maintenance and provide evidence that aerobic glycolysis is part of a metabolic program that supports the biosynthetic needs of a normal neuronal cell type.

https://doi.org/10.7554/eLife.25946.001

eLife digest

Living cells need building materials and energy to grow and carry out their activities. Most cells in the body use sugars like glucose for these purposes. In a process known as glycolysis, cells break down glucose into molecules that are eventually converted to carbon dioxide and water to form the chemical ATP – the cellular currency for energy. Developing cells that have not yet fully specialized, and rapidly dividing cells, like cancer cells, consume large amounts of glucose via aerobic glycolysis (also known as the Warburg effect) as they require high levels of energy and building materials. As cells become more specialized and divide less often, they have a reduced need for building blocks, and adjust their consumption and breakdown of glucose accordingly.

One exception is the photoreceptor cells, found in the light-sensitive part of our eyes. Although these specialized cells do not divide, they still need a lot of energy and building blocks to constantly renew their light-sensing and processing structures, and to capture and convert the information from the environment into signals. Previous research has shown that the eye also uses the Warburg effect. However, until now, it was not known whether the photoreceptors or other cells in the eye carry out this form of glycolysis.

Using genetic tools, Chinchore et al. analysed how the photoreceptor cells in mice used glucose. The experiments demonstrated that the photoreceptors do indeed carry out the Warburg effect. Chinchore et al. further discovered that the Warburg effect is regulated by the same key enzymes and signalling molecules that cancer cells use. This indicates that specialized cells like photoreceptors might choose to retain certain metabolic features of their precursor cells, if they need to.

These findings provide new insight into how photoreceptors use glucose. The next step will be to understand how aerobic glycolysis is regulated in healthy eyes as well as in eyes that are affected by degenerating diseases, which may ultimately lead to new ways of treating blindness.

https://doi.org/10.7554/eLife.25946.002

Introduction

Sensory neurons capture information from the environment and convert it into signals that can greatly impact the survival of an organism. These systems are thus under heavy selective pressure, including for the most efficient use of energy to support their sensitivity and efficiency (Niven and Laughlin, 2008). In this regard, the primary photoreceptor cells face a dual challenge. They not only need to preserve their membrane excitability via ion pumps by ATP hydrolysis (Okawa et al., 2008) but also maintain elaborate membranous organelles (rhabdomeres in invertebrates and outer segments in vertebrates) that maximize light capture. The maintenance of such structures requires considerable metabolic resources. Invertebrate photoreceptors exhibit light-dependent endocytosis of their photosensitive membranes (Satoh et al., 2005) enabling the recycling of these resources. In contrast, vertebrate photoreceptors shed a fraction of their outer segments (OS) daily, to be phagocytosed by the juxtaposed retinal pigmented epithelium (RPE) (Basinger et al., 1976; LaVail, 1976) (Figure 1—figure supplement 1). To sustain a constant volume of the OS, a cell must channel metabolites toward biosynthesis, against the backdrop of very high ATP consumption, which is required to maintain membrane potential. Photoreceptors thus must balance the use of their intracellular carbon pool between oxidative catabolism, to generate the required ATP, and anabolism, to continually renew the OS.

The mammalian retina depends upon glucose and glycolysis for survival and function (Chertov et al., 2011; Noell, 1951). The majority (~80%) of glucose is converted to lactate via glycolysis (Cohen and Noell, 1960; Warburg, 1925; Winkler, 1981). The adult retina can produce lactate aerobically (aerobic glycolysis/Warburg effect) with an ~50% increase during anaerobic conditions (Pasteur effect) (Cohen and Noell, 1960). The cell types that carry out aerobic glycolysis in the normal adult retina have not been determined. The photoreceptors have been assumed to rely on aerobic glycolysis. This assumption is based on the adverse effects on photoreceptor function after en masse inhibition of whole retinal glycolysis by pharmacological treatments e.g. with iodoacetate (Winkler, 1981). The Warburg effect exemplifies an elaborate set of metabolic strategies adopted by a cell to preferentially promote glycolysis (Gatenby and Gillies, 2004; Liberti and Locasale, 2016). One drawback of inhibiting glycolytic enzyme activity in the retina is that such a manipulation does not differentiate between aerobic glycolysis and housekeeping glycolysis- a pathway critical for most cell types.

Studies conducted on retinal tissue in vitro indicate that isolated mammalian photoreceptors can consume lactate, which can be produced by glycolysis in retinal Mueller glia (Poitry-Yamate et al., 1995). Thus, the decreased photoreceptor function after whole retinal glycolytic enzyme inhibition could be a non-cell-autonomous effect on Muller glia. Although many features of the ‘lactate shuttle’ and its in vivo relevance have recently been questioned (Hurley et al., 2015), it is important to devise an experimental strategy that would be able to discern the cell-autonomous versus non-autonomous requirement of glycolysis for the photoreceptors.

The cellular origins and purpose of aerobic glycolysis in the retina, its relevance to photoreceptor physiology, and its regulation, are not understood. In this study, we explored the propensity of photoreceptors to produce or consume lactate and utilized genetic manipulations to reveal the regulatory mechanisms of glycolysis. We show that rod photoreceptors rely on glycolysis for their OS biogenesis. Genetic perturbations targeting allostery or key regulatory nodes in the glycolytic pathway impacted the OS size. Fibroblast growth factor (FGF) signaling was found to regulate glycolysis, with antagonism of this pathway resulting in anabolic deficits. These data demonstrate the cell autonomous role of the glycolytic pathway in OS maintenance and provide evidence that aerobic glycolysis is part of a metabolic program that supports the biosynthetic needs of a normal neuronal cell type.

Results

Aerobic glycolysis in the retina

We first examined lactate production from the retina and assayed the metabolic consequences of inhibiting aerobic glycolysis. Lactate is produced by reduction of pyruvate, a reaction catalyzed by lactate dehydrogenase (LDH) (Figure 1—figure supplement 2A). Freshly isolated retinae were cultured in the presence or absence of sodium oxamate- an LDH inhibitor. These were subsequently transferred to buffered Krebs’-Ringer’s medium that has glucose as the sole source of carbon (see - an LDH inhibitor. These were subsequently transferred to buffered Krebs’-Ringer’s medium that has glucose as the sole source of carbon (see Materials and methods), and lactate secretion was quantified (Figure 1A). The extracellular secreted lactate was measured because it represents the pyruvate-derived carbons that are diverted away from other intracellular metabolic processes or the mitochondria. Oxamate treatment led to a significant drop in the secreted lactate production rate compared to control. In addition, the ATP levels were monitored and, surprisingly, the steady-state levels of ATP in oxamate-treated retinae did not differ from the control retinae (Figure 1B). This could be due to a relatively minor glycolytic contribution to the total ATP pool, a compensatory metabolic realignment toward mitochondria-dependent ATP production or existence of phosphotransfer enzyme systems such as adenylate kinase or creatine kinase. To differentiate among these possibilities, explants were cultured in oxamate or control conditions followed by a short treatment with NaN3 to inhibit mitochondrial ATP synthesis (Figure 1B). Control retinae displayed ~50% reduction in ATP levels after incubation in NaN3. Interestingly, oxamate-treated retinae displayed a further 20% decrease in ATP after exposure to NaN3. Thus, inhibiting lactate synthesis resulted in a greater fraction of the ATP pool that was sensitive to mitochondrial function.

Figure 1 with 4 supplements see all
Ldha-dependent aerobic glycolysis and outer segment maintenance in photoreceptors.

(A) Freshly explanted retinas were treated with the LDH inhibitor, sodium oxamate, for 8 hr in explant culture medium, transferred to Krebs’-Ringer's for 30 min, and lactate was measured in the supernatant. Control (n = 5), Oxamate (n = 4). (B) Freshly explanted retinas were treated with oxamate or NaCl (control) in explant culture medium for 8 hr, followed by treatment with NaN3 or NaCl (untreated group) in Krebs’-Ringer's medium for 30 min. ATP per retina was then measured. n = 7, Control untreated; n = 8, Oxamate untreated n = 8, Control NaN3; n = 8, Oxamate NaN3. (C) Expression of Ldha and Ldhb as determined by IHC. Glutamine synthetase (GS), a Mueller glia-specific marker, colocalized with LDHB in the cell bodies (arrowheads), processes ensheathing the photoreceptors (arrows) and the outer limiting membrane (OLM, *). Scale bar, 50 μm. (D) ISH for Ldha and Ldhb. Ldha RNA displayed photoreceptor-enriched expression while Ldhb RNA was not observed in photoreceptors. Scale bar, 100 μm. (E, F) Freshly explanted retinas were treated with FX11 or DMSO for 8 hr and transferred to Krebs’-Ringer's for 30 min and secreted lactate was measured (E) n = 5, DMSO; n = 6, FX11, or they were transferred to Krebs’-Ringer's buffer with NaN3 or NaCl (untreated group) for 30 min for ATP quantitation (F). ATP per retina was measured at the end of the assay. n = 8, DMSO untreated; n = 8, FX11 untreated; n = 9, DMSO NaN3; n = 7, FX11 NaN3. (G) Freshly explanted retinae were transferred to Krebs’-Ringer's for 30 min and secreted lactate was measured. n = 8, Bl6/J; n = 8, Ldhafl/fl; n = 8, Rod-cre; n = 16, Rod-cre> Ldhafl/fl; n = 8, Rod-cre> Ldhafl/+. (H) Photoreceptor outer segment phenotype 42–45 days following in vivo electroporation of a knock-down construct (shRNA) for Ldha. CAG-mGFP was used for coelectroporation. Plasmid combinations listed on the left. Magnification of areas outlined in yellow is displayed on right with threshold-adjusted rendering to highlight inner and outer segments. Scale bar, 25 μm. (I) Quantification of inner+outer segment (IS+OS) lengths. n = 53–74 photoreceptors, 4–5 retinae. (J) Photoreceptor outer segment phenotype of dark-reared animals. Electroporated pups were transferred to dark on the day of eye opening (P11) and reared with their mothers for 3 weeks. (K) Quantification of inner+outer segment lengths of (J). n = 53–83 photoreceptors, 4–5 retinae. (L) Colored end products of redox reactions catalyzed by COX and SDH enzymes in retinal tissue. Scale bar, 200 μm. (M) IHC for SDH-A subunit in adult retina. Scale bar, 200 μm. ONL, outer nuclear layer. INL, inner nuclear layer. Data, Mean±SD. Statistics, unpaired, two-tailed t-test with Kolmogorov-Smirnov correction for panels A, E; two-way ANOVA with Tukey’s correction for panels B, F and K; one-way ANOVA with Tukey’s multiple comparison test for panels G, I.

https://doi.org/10.7554/eLife.25946.003

Lactate producing isoform of LDH in photoreceptors

Next, we wanted to ascertain if photoreceptors produce or consume lactate. As a first step, the expression of the LDH subtypes was examined. LDH is a tetrameric enzyme composed of LDHA and LDHB subunits encoded by the Ldha and Ldhb genes respectively. The subunits can assemble in five different combinations with differing kinetic properties (Dawson et al., 1964; Doherty and Cleveland, 2013) (Figure 1—figure supplement 2B). A tetramer of all LDHA subunits has high affinity for pyruvate and a higher Vmax for pyruvate conversion to lactate than does an all-LDHB isoenzyme. In addition, many glycolytic cancers have elevated Ldha expression (Balinsky et al., 1983; Behringer et al., 2003). On the contrary, an all-LDHB tetramer is maximally active at low pyruvate concentrations, is strongly inhibited by pyruvate, and is expressed in tissues using lactate for oxidative metabolism or gluconeogenesis (Dawson et al., 1964). We examined the expression of LDHA and LDHB subunits in the retina by immunohistochemistry (IHC) (Figure 1C). Photoreceptors showed strong expression of LDHA, particularly with respect to the other retinal cell types. Similar results were obtained by others using a different set of commercially available antibodies (Casson et al., 2016; Rueda et al., 2016). Immunohistochemical localization also indicated that LDHB was abundantly expressed in the cells of the inner nuclear layer (INL), which includes interneurons and Mueller glia. To validate the staining pattern obtained by IHC, expression analysis of transcripts of Ldha and Ldhb genes by in situ hybridization (ISH) was performed (Figure 1D). This confirmed that Ldha expression is enriched in the photoreceptors, whereas Ldhb is excluded. This was also confirmed by qRT-PCR analysis of Ldha and Ldhb transcripts in isolated rod photoreceptor cDNA (Supplementary file 1). We conclude that photoreceptors have predominantly LDHA-type subunits.

We also assessed the ability of the retina to secrete lactate after treatment with the LDHA-specific inhibitor, FX-11 (Le et al., 2010). FX-11 significantly reduced lactate secretion (Figure 1E). Similar to oxamate, FX-11 also resulted in an increased percentage of ATP that was sensitive to azide inhibition (Figure 1F). To investigate if photoreceptors produce lactate in an Ldha-dependent manner, mice with a conditional allele of Ldha (Wang et al., 2014b), (Ldhafl/fl), were used. The specificity and efficiency of Cre recombinase under control of the rhodopsin regulatory elements (Le et al., 2006) were first tested, which showed that only rod photoreceptors had a history of cre expression (Figure 1—figure supplement 2C,D) (the mouse line henceforth called Rod-cre). The recombination efficiency varied between ~50–90% of photoreceptors among different retinae. The Rod-cre; Ldhafl/fl retinae were examined for LDHA protein, which showed a significant reduction (Figure 1—figure supplement 2E). A compensatory expression of Ldhb in photoreceptors was not detected (Figure 1—figure supplement 2F). Lactate production in these retinae was examined and was found to be significantly reduced (Figure 1G). Thus photoreceptors produce lactate in an Ldha-dependent manner.

Active LDHA supports outer segment biogenesis

In order to assess if reduction in Ldha expression created a cellular phenotype in photoreceptors, and if so, whether it was required autonomously, electroporation of a short hairpin RNA (shRNA) specifically targeting the 3’ untranslated region (UTR) of the Ldha transcript was used (Figure 1—figure supplement 2G). This strategy was taken, vs.examination of the rods in the Rod-cre; Ldhafl/fl retinae, due to the concern that a reduction in lactate by rods might affect closely associated cell types, such as Mueller glia and/or RPE cells, creating non-autonomous effects on rods. A plasmid encoding this shRNA was delivered to the retina in vivo by electroporation. Electroporation occurs in patches comprising 15–30% of the retina, and in a given patch, only ~20% cells are electroporated (Sui Wang and C. Cepko, unpublished). Thus, plasmid transfection via electroporation allowed us to determine if Ldha has a cell-autonomous role in photoreceptors. The electroporated photoreceptors had markedly reduced OS length when compared to control (Figure 1H,I). Genetic complementation by coelectroporation of a sh-resistant Ldha cDNA that lacks the 3’UTR demonstrated that the defect was attributable to Ldha loss-of-function (Figure 1H,I) and the phenotype observed with the shRNA was not due to off-target effects. To determine if the catalytic activity of LDHA was required for rescue, an allele of Ldha with a point mutation in the catalytic center (LDHAH193>A) was introduced. It failed to rescue the shRNA phenotype. Finally, expression of Ldhb was not sufficient to compensate for Ldha loss-of-function (Figure 1—figure supplement 2H). To confirm that the Ldha knockdown via electroporation conferred a cell-autonomous phenotype, we examined rhodopsin localization in mGFP-negative rods within the electroporated patch (Figure 1—figure supplement 3). A non-autonomous deleterious effect on rods that did not receive the plasmid (mGFP-) was not observed. The rhodopsin localization and the length of the OSs in GFP-negative rods within the patch did not vary from that of the rods lying outside of the electroporated patch (Figure 1—figure supplement 3).

The cyclical process of OS shedding and renewal is regulated by light (LaVail, 1976). Since LDHA function is necessary to maintain OS length, we assessed the effect of Ldha knockdown in dark-reared mice and compared with mice raised in normal room light. Electroporated mice were raised with their mothers in normal room light until eyes were open (P11), and then shifted to the dark for 3 weeks. In mice with no Ldha knockdown, there was ~25% increase in IS+OS length after dark rearing compared to the light:dark condition (Figure 1J,K), presumably as a part of an adaptive mechanism that might include less OS shedding (Penn and Williams, 1986). Interestingly, in mice with the Ldha knockdown, dark rearing resulted in a partial rescue of the Ldha knockdown phenotype (Figure 1J,K). The average IS+OS length after Ldha knockdown was similar to that of light-treated control animals. These data indicate that reducing the need for OS biogenesis, as occurs in the dark, led to a reduced reliance on Ldha function.

Functional mitochondria in photoreceptors

Cells with immature or dysfunctional mitochondria become reliant on glycolysis by increasing Ldha expression at the expense of Ldhb (Facucho-Oliveira et al., 2007; Ross et al., 2010; Trifunovic et al., 2004). Although photoreceptors have abundant mitochondria, a reason for their high Ldha and low Ldhb expression could be subpar mitochondrial function, especially when compared to other retinal cell types. Thus, we assessed whether there was a mitochondrial activity difference between the photoreceptors and INL cells by examining succinate dehydrogenase (SDH) and cytochrome oxidase (COX) activity in fresh, unfixed, adult retinal sections (Figure 1L). SDH/complex II plays a role in the citric acid cycle, as well as in the electron transport chain, and its subunits are encoded by the nucleus. COX or complex IV plays a role in the electron transport chain and has catalytic subunits that are encoded by the mitochondrial genome (mtDNA). SDH activity was not lower in the photoreceptors relative to INL cells. COX activity was high in the photoreceptor layer, even higher than that seen in the other retinal layers. The specificity controls for the histochemical reaction are presented in Figure 1—figure supplement 2I. Finally, IHC for SDH was carried out. The highest IHC signal was observed in the photoreceptor inner segments (IS), as well as the OPL and IPL synaptic layers (Figure 1M), in good agreement with the observed SDH activity. IHC for another mitochondria-specific enzyme, pyruvate dehydrogenase, showed a similar pattern (Figure 1—figure supplement 2J) indicating that these are the sites of maximal mitochondrial densities in the retina. These data align with other studies that assessed mitochondrial activity in the retina (Hansson, 1970; Rueda et al., 2016). Thus, lactate production by the photoreceptors cannot be attributed to lack of mitochondrial activity. Similarly, a decrease in mitochondrial COX activity was not detectable in the Rod-cre; Ldhafl/fl retinae (Figure 1—figure supplement 4) by the histochemical assay.

Allosteric regulation of glycolysis in photoreceptors

LDHA supports glycolysis by providing a ready supply of cytosolic NAD+ that is independent of O2 availability and/or mitochondrial function. The phenotype observed following Ldha knockdown might be indicative of a reliance on glycolysis where cells might exhibit a preference for unabated and rapid flux through glycolysis. Alternatively, it could be due to an unidentified role of Ldha in OS maintenance. To understand the extent of photoreceptors’ dependence on glycolysis, we designed an experimental strategy that satisfied the following criteria: (1) Does not ablate core glycolytic enzymes in order to avoid pleiotropic effects due to their possible non-glycolytic roles, (2) Targets a glycolytic node such that impact on other biosynthetic pathways, such as Pentose Phosphate Pathway (PPP), would be minimal and (3) Uncovers glycolytic reliance and differentiates it from ‘housekeeping’ glycolysis. Glucose-derived metabolites are committed towards glycolytic flux by the enzyme 6-phosphofructo-1-kinase (PFK1), which catalyzes conversion of fructose-6-phosphate (F6P) to fructose-1,6-bisphosphate (F-1,6-BP) (Figure 2—figure supplement 1.). The most potent allosteric activator of PFK1 is fructose-2,6-bisphosphate (F-2,6-BP) (Hers and Van Schaftingen, 1982). F-2,6-BP is synthesized from F6P by the kinase activity of the bifunctional enzyme, 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFK2) (Figure 2—figure supplement 1A,B). To examine the glycolytic dependence of photoreceptors, we targeted the steady-state levels of the metabolite, F-2,6-BP as it would satisfy the above criteria.

First, we examined expression of PFK2 isoenzymes encoded by Pfkfb1-4 genes (Figure 2—figure supplement 1C). Pfkfb3 expression could not be detected. Pfkfb1, 2 and 4 were expressed in either a photoreceptor-enriched or photoreceptor-specific pattern, suggesting a propensity of these cell types to regulate glycolysis via a PFK2-dependent mechanism.

With the exception of Pfkfb3, all other PFK2 isoenzymes have kinase and phosphatase domains on the same polypeptide (Mor et al., 2011) (Figure 2—figure supplement 1B). In addition to potential problems posed by functional redundancy (i.e. knockdown of one isoenzyme might not be sufficient), genetically ablating the PFK2 isoforms would not uncover the preference for directionality (i.e. an observed phenotype could be attributed to absence of either the kinase or phosphatase function). In addition, the structure-function relationships of their kinase and phosphatase domains are not known, thus making kinase- or phosphatase-dead versions is not straightforward. To overcome these problems, we overexpressed Tigar (TP53-induced glycolysis and apoptosis regulator) as it is functionally similar to the phosphatase domain of PFK2 with well-characterized F-2,6-BPase activity (Bensaad et al., 2006) (Figure 2—figure supplement 1B) and hence reduces the steady state levels F-2,6-BP. This is the intended effect and bypasses the aforementioned concerns with the conventional genetic loss-of-function approach associated with PFK2 isoenzymes. In addition to the predicted function of reducing glycolysis, overexpression of Tigar would not negatively affect the PPP (Bensaad et al., 2006).

We utilized an experimental strategy that addressed the following concerns: (1) The effect should be autonomous to photoreceptors, (2) the phenotype should be induced in fully mature photoreceptors, and (3) the phenotype should discernably be due to perturbations specifically of fructose-2,6-bisphosphate. Our experimental scheme utilized a construct that expressed tamoxifen-inducible Cre only in rods (Figure 2A,B and Figure 2—figure supplement 1D). Expression of Tigar specifically in adult photoreceptors resulted in a significant reduction of OS length (Figure 2C,D). This phenotype was specifically attributable to the phosphatase activity because expression of a catalytic dead version of Tigar, Tigar-TM (triple mutant, H11>A, E102>A, H198>A)(Bensaad et al., 2006), did not cause a change in the photoreceptor OS length (Figure 2—figure supplement 1E,F). To ascertain if the phenotype is specifically attributable to Tigar’s phosphatase activity on F-2,6-BP, we decided to coexpress Pfkfb3- a PFK2 isoform that has the kinase activity ~700 fold higher than the phosphatase (Sakakibara et al., 1997) (Figure 2A and Figure 2—figure supplement 1B). Interestingly, overexpression of Pfkfb3 alone did not result in an overt phenotype- the OS length and morphology were indistinguishable from those of the control electroporated retina (Figure 2C,D). Overexpression of Pfkfb3 was able to rescue the reduction in OS length caused by Tigar expression (Figure 2C,D). Together, these data suggest that adult photoreceptors are sensitive to perturbations targeting F-2,6-BP.

Figure 2 with 1 supplement see all
Targeting allostery reveals glycolytic reliance for outer segment maintenance.

(A) Constructs for spatio-temporal control of expression of Tigar and Pfkfb3. DsRed used as the cre reporter, mGFP as a coelectroporation marker. (B) Scheme for electroporation and tamoxifen induction. i.p, Intraperitoneal. (C, D) IS+OS length were measured following introduction of Tigar (n = 72 cells), PFKB3 (n = 72 cells), and Tigar and Pfkfb3 constructs (n = 78), shown in (A). Controls were -tamoxifen (n = 62) and +tamoxifen (n = 74). Scale bar, 50 μm. Data are Mean±SD. One-way ANOVA with Tukey’s correction. Outlined areas magnified to show IS and OS morphology. (E) AAV genomes for expression of mGFP (AAV-mGFP) or Tigar (AAV-TIGAR). (F) Cross-sections of AAV-mGFP alone or AAV-TIGAR (coinjected with AAV-mGFP) infected retinae harvested at P28 imaged for mGFP expression. (G) Intracellular lactate normalized for total protein was quantified for retinae infected with AAV-mGFP (Control) or AAV-mGFP + AAV-TIGAR. Data represented as percentages relative to age-matched, freshly isolated retinae. Data, Mean±SD. Unpaired, two-tailed t-test with Kolmogorov-Smirnov correction for non-Gaussian distribution. ONL, outer nuclear layer.

https://doi.org/10.7554/eLife.25946.009

Next, the effects of Tigar expression on glycolysis were assayed. Although electroporation answers the question of the cell autonomous effect of a perturbation, the total number of affected cells and their percentage in the electroporated area are too minor to determine biochemical contributions. Adeno-associated virus (AAV)-mediated transduction of Tigar into photoreceptors was thus used, as it transduces a greater percentage of cells than electroporation. An AAV construct that drives expression of Tigar and/or mGFP from the bovine rhodopsin (RHO) promoter specifically in rods was constructed (Figure 2E,F and Figure 2—figure supplement 1G). The AAVs (expressing mGFP alone or mGFP and TIGAR) were injected at postnatal day 6 (P6), after the end of cell proliferation, and nearly full retinal infection was confirmed by indirect ophthalmoscopy at P24-P27 by assessing the GFP fluorescence. Retinae were harvested at P28 and examined for expression (Figure 2—figure supplement 1H) and lactate levels (Figure 2G). Consistent with the idea that Tigar would interfere in allosteric regulation of glycolysis, a significant reduction in retinal lactate was observed in the AAV-TIGAR infected retinae compared to the control AAV-mGFP infected retinae (Figure 2G).

Nonequivalent roles of pyruvate kinase isoforms

Given the essential role of Ldha in postmitotic photoreceptors and proliferating cancer cells, other aspects of metabolism that have been discovered in cancer cells, such as the expression of pyruvate kinase isoforms, were investigated. Pyruvate kinase catalyzes the final irreversible reaction of glycolysis and distinct isoenzymes are encoded by two genomic loci, Pkm (muscle) and Pklr (liver and red blood cell). Pklr transcripts were not detected in the retina, (Figure 3—figure supplement 1), but M1 and M2 splice isoforms of the PKM gene were detected (Figure 3—figure supplement 2A) in line with protein expression data reported earlier (Lindsay et al., 2014). The M2 isoform is known to regulate aerobic glycolysis, promotes lactate production, and is upregulated in many tumors (Christofk et al., 2008a, 2008b). This isoform was previously reported to be expressed in photoreceptors (Casson et al., 2016; Lindsay et al., 2014; Morohoshi et al., 2012; Rajala et al., 2016; Rueda et al., 2016). We confirmed that there is photoreceptor-enriched expression of PKM2 by IHC (Figure 3A). PKM1, known to be expressed in most differentiated cell types in adults (Jurica et al., 1998), was expressed in the cells of the INL and ganglion cell layer, as shown by IHC (Figure 3A), but was not detectable in photoreceptor cells. In this regard, our data differed from some published findings (Casson et al., 2016; Lindsay et al., 2014) but matched those of others (Rajala et al., 2016). To address this discrepancy and validate commercially available antibodies, we performed isoform-specific ISH (Figure 3A) and confirmed the expression pattern that we observed using IHC. We also examined transcript abundance by qPCR in mRNA purified from isolated rod photoreceptor cells (Supplementary file 1) and found the M1 isoform to be much less abundant than M2 in the photoreceptors.

Postnatally, PKM1 protein expression gradually increased, in correlation with increased differentiation and decreased proliferation in the developing retina (Figure 3B). On the other hand, PKM2 protein expression was detectable during the period of proliferation and its expression did not decrease with increased differentiation, likely due to retention of expression in differentiated photoreceptors. Previous studies on pyruvate kinase in the context of proliferation have suggested that loss-of-function of Pkm2 reduces proliferation attributable to the glycolytic reliance of mitotic cells for growth (Christofk et al., 2008b; Israelsen et al., 2013). To assess if PKM2 plays an essential role in rod photoreceptors, an shRNA construct that specifically targeted mouse PKM2 (PKM2sh), but spared PKM1 (Figure 3—figure supplement 2B,C,D,E), was generated. In vivo electroporation of a plasmid encoding PKM2-specific shRNA resulted in photoreceptors with significantly shorter OS than control (Figure 3D,E). This phenotype could be rescued by coelectroporation of a construct encoding human PKM2 cDNA (Figure 3C,D), which was not targetable by the shRNA (Figure 3—figure supplement 2F). Coelectroporation of plasmid encoding mouse PKM1 with PKM2sh did not rescue the OS length defect (Figure 3C,D). These data demonstrate that PKM1 and PKM2 play nonequivalent roles in rod photoreceptors. In order to further investigate whether PKM2 was needed for an autonomous role in rods, we electroporated a low concentration of plasmid encoding the shRNA (Figure 3—figure supplement 3). In addition to few electroporated rods, there were very few electroporated INL cells. In this condition, electroporated rods displayed a similar OS phenotype to that observed with higher concentrations of PKM2sh (Figure 3C), that is, reduced OSs. We also generated an shRNA construct that targeted exon 4, which is shared between mouse PKM1 and PKM2 (PKM1 +2 sh) (Figure 3—figure supplement 2C,D). Electroporation of this construct resulted in a significant decrease in the OS length (Figure 3—figure supplement 2G). The photoreceptor morphology and OS length were the same as that observed following electroporation with PKM2sh. While complementation with human PKM2 was sufficient to rescue the IS+OS length defect, we noted some abnormalities with the morphology of some of the photoreceptor ISs and OSs (Figure 3—figure supplement 2G). In 4/6 retinae, many photoreceptors lacked clear borders of IS and OS, though in 2/6 retinae, the morphology closely resembled that of control retinae (Figure 3—figure supplement 2G).

The contribution of PKM2 to OS maintenance was further investigated in the retinae of dark-reared mice electroporated with PKM2sh (Figure 3E). Dark rearing significantly increased OS length in these animals (Figure 3F). Taken together, the results from dark-reared animals, in which Ldha or Pkm2 was knocked down, indicate the requirement for the glycolytic pathway in OS maintenance. Since two different genes that promote aerobic glycolysis are necessary for the light-dependent maintenance of OS, the short OS phenotype is likely due to a reduced supply of the building blocks normally supplied by aerobic glycolysis.

In order to probe the biochemical effects of PKM2 reduction, lactate production was examined. Since electroporated retinae are not ideal for these experiments, mice that had a conditional deletion of Pkm2 in rods were used. The Pkm2fl/fl mouse strain, in which the M2-specific exon 10 was floxed (Israelsen et al., 2013), was crossed with the Rod-cre strain. The retinae with deficiency of PKM2 had a small but significant decrease in lactate production, as compared to the controls (Figure 3G). We also noted upregulation of PKM1 in these retinae (Figure 3—figure supplement 2H) similar to what has been reported before (Israelsen et al., 2013). However, as noted above, in rods electroporated with PHM2sh, PKM1 expression was not observed (Figure 3—figure supplement 2I). One possibility for the difference in the presence of the M1-specific exon in the mRNA in the knockout vs. the knockdown manipulation might reflect a choice made by the splicing machinery. After the deletion of the ‘preferred’ M2-specific exon in the genome in the knockout, the splicing machinery might include the M1 exon as a default choice. However, when the shRNA was used to knockdown the PKM2 isoform, the splicing event that chose the M2-specific exon would have already happened.

Figure 3 with 6 supplements see all
PKM1 and PKM2 isoforms have nonequivalent roles.

(A) Biased expression of M1 and M2 isoforms in retinal layers detected by IHC and ISH. (B) Immunoblot of retinal lysates from postnatal retina at different developmental stages. HEK293T cell lysates that were from untransfected (UT) cells, or those transfected with CAG-FLAGmuPKM1 (M1) or CAG-FLAGmuPKM2 (M2) as controls. Postnatal age in days. A, mature retina (P25–P30). (C) Outer segment phenotype of P45 mice after electroporation with constructs encoding mouse PKM2-specific shRNA (PKM2sh) and adding either mouse PKM1 (muPKM1) or human PKM2 (huPKM2). Selected areas in yellow boxes are magnified on the right. (D) Quantification of IS+OS lengths obtained in (C). n = 32–53 cells from 3 to 4 retinae. (E) Outer segment phenotype of dark-reared P31 mice electroporated with PKM2sh-encoding plasmid. The yellow-boxed region is magnified and presented on the right. (F) Quantification of IS+OS lengths obtained in (e). n = 75 cells from three retinae. (G) Secreted lactate from freshly isolated retinae from Pkm2fl/fl (fl/fl) (n = 12) or Rod-cre> Pkm2fl/fl (m2-/-) (n = 16) mice. (H) Outer segment phenotype after CAG promoter-driven overexpression of Flag-tagged mouse PKM1 or PKM2. Inset, higher magnification of IS and OS. (I) Quantification of IS+OS lengths obtained in (H). n = 35 cells from three retinae in PKM1 and PKM2 groups. ONL, outer nuclear layer. Data, Mean±SD. Statistics, one-way ANOVA with Tukey’s correction for panels D, I; two-way ANOVA with Tukey’s multiple comparison test for panel F; unpaired, two-tailed t-test with Kolmogorov-Smirnov correction for panel G.

https://doi.org/10.7554/eLife.25946.012

The differential expression of the M1 and M2 isoforms in the retinal layers could be attributable to the differential expression of splicing factors that promote inclusion or exclusion of the M1- or M2-specific exon. To evaluate this possibility, we examined the expression of Srsf3, a splicing factor known to promote inclusion of the M2 exon (Wang et al., 2012), and Ptbp1, known to repress the M1 exon inclusion (Chen et al., 2012) (Figure 3—figure supplement 4). While Srsf3 was expressed at higher levels in photoreceptors, Ptbp1 was more enriched in the INL. Thus the regulation of PKM isoform preferences in retina is more complex than that predicted by canonical splicing models.

We also noted slightly reduced rod OS length in the region that had PKM1 expression in the young (postnatal 6 week) Rod-cre; Pkm2fl/fl mice (Figure 3—figure supplement 5). Although the recombination in this line has been reported to be complete by 6 weeks, we cannot exclude the possibility of some recently recombined rods (that express PKM1) in this region. These rods might not have had enough time to have a discernable impact on rhodopsin abundance and OS length. In addition, there might be some non-recombined rods interspersed in the broad region where PKM1 expression was apparent, and contributed to longer OSs. Due to the packing density of rods and abundance of rhodopsin, an immunohistochemical approach might not be a suitable way to reliably assess the OS length and its relation to PKM2 function. We also examined older mice (37 weeks, ~8 months old) of this background hypothesizing that aging might uncover a subtle phenotype (Figure 3—figure supplement 6). The nuclei in the ONL region that had lost PKM2 protein expression were disorganized and lost their typical columnar arrangement, perhaps due to cell death. We also noted disorganization of OS in some regions where PKM2 was lost. Overall the OS length was reduced compared to aged Bl6/J mice, but this reduction was also noted for photoreceptors that had PKM2 protein expression. We cannot exclude the possibility of a non-autonomous effect on these rods as a response to tissue reorganization or alterations in their metabolic environment. Similarly, maintenance of some rhodopsin expression after PKM2 loss in the surviving rods might be an indication of an adaptive response on part of these cells. The electroporation approach, where only a few cells are transfected, circumvents these concerns and illustrates the critical cell-autonomous requirement of PKM2 for rods.

PKM1 is constitutively active while PKM2 is regulatable (Anastasiou et al., 2012). Biased expression of PKM2 in photoreceptors suggests that these cells may need to dynamically regulate glycolysis. The inability of PKM1 to rescue PKM2 loss-of-function indicates that merely replacing pyruvate kinase (PK) function after Pkm2 knockdown is not sufficient to restore the OS. In addition, it indicates the importance of glycolytic regulation at the PK step in photoreceptors. We examined the effect of forced expression of PKM1 in the presence of endogenous PKM2, with the hypothesis that the constitutively active isoform might interfere at the regulatory step. We delivered plasmids encoding FLAG-tagged mouse PKM1 and PKM2 via in vivo electroporation (Figure 3H). Photoreceptors electroporated with PKM1-expressing constructs, but not PKM2 expressing constructs, had a reduction in the length of the OS (Figure 3H,I) with the majority of the photoreceptors in the PKM1 electroporated retinae lacking discernable OS. The two proteins were expressed at equivalent levels, as assessed by Western blotting for the FLAG epitope in HEK293T cells (Figure 3—figure supplement 2J).

Fibroblast growth factor signaling regulates anabolism

PKM2 has been shown to interact with tyrosine phosphorylated proteins (Christofk et al., 2008a) and is tyrosine phosphorylated at position 105 (pY105) in tumor cells (Hitosugi et al., 2009) leading to promotion of aerobic glycolysis. The pY105 is a shared epitope in PKM1 and PKM2 (Figure 4A). To assess the phosphorylation status of PKM2 at this site, PKM2 was specifically immunoprecipitated from retinal lysates followed by immunoblotting using a phospho-Y105-specific antibody (Figure 4B). We observed that PKM2 was phosphorylated at Y105. In order to ascertain if phosphorylation of PKM2 at this site might have any physiological significance, its regulation by light was examined. PKM2 was immunoprecipitated from the retinae of mice at 3 hr intervals during a 24-hr time course, and phosphorylation at Y105 was probed (Figure 4C). A light-dependent increase in phosphorylation at Y105 of PKM2 was observed. Thus, this phosphorylation site might be one of the target sites for physiologically relevant signaling events regulating aerobic glycolysis. Light-dependent phosphorylation of this epitope in the retina has also been reported recently using IHC and immunoblot approaches (Rajala et al., 2016). The authors observed changes in signal intensity on IHC, that were dependent on light and activation status of the phototransduction pathway. Similarly, the Y105 epitope showed light-dependent phosphorylation as assessed by immunoblot analysis of total retinal lysates. Our results on immunoprecipitated PKM2 confirm that this protein is among the targets of a light-dependent signaling pathway. Thus, phosphorylation of this site was then used as a proxy for the tyrosine kinase signaling pathways that could phosphorylate PKM2 in the retina. Freshly explanted retinae were cultured with antagonists targeting specific pathways: Afatinib (EGFR), Dasatinib (Src), BMS536924 (Insulin/IGF), PD173074 (FGFR1) and Dovitinib/TKI258 (FGFR1 and FGFR3). PKM2 was immunoprecipitated and its phosphorylation at Y105 probed (Figure 4D). FGF inhibitors, PD173074 and TKI258, reduced PKM2 phosphorylation. Tyrosine kinase signaling can also target multiple nodes, including pyruvate dehydrogenase kinase and LDHA (Fan et al., 2011), and regulate aerobic glycolysis in cancer. We observed that treatment with either PD173074 or TK1258 also resulted in a dose-dependent decrease in LDHA phosphorylation at the Y10 residue (Figure 4E). Thus, FGF signaling potentially targets multiple nodes in order to regulate aerobic glycolysis in the retina.

Figure 4 with 1 supplement see all
FGF signaling regulates aerobic glycolysis and anabolism.

(A) Schematic of PKM1 and PKM2 polypeptide showing Y105 is a shared epitope between PKM1 and PKM2. (B) Immunoprecipitation (IP) of PKM2 from adult retina followed by immunoblot (IB) for either PKM1, PKM2 or pY105 PKM. IP using isotype-matched antibody (IgG) is used alongside to control for nonspecific binding. Lysates from skeletal muscle (expresses PKM1) and 293T (expresses only PKM2) included as controls. Molecular weight marker positions are depicted on the right-hand-side (C) Retinal lysates were prepared from eyes harvested at 3-hr interval during the 12 hr light 12 hr dark cycle. T0 is the time point of light on in the room. The lysates were subjected to immunoprecipitation with anti-PKM2. Immunoprecipitates were probed for phosphorylation at Y105 by immunoblotting with the phospho-specific antibody. (D) Lysates from explants treated with candidate tyrosine kinase pathway inhibitors or vehicle control (DMSO) were subjected to immunoprecipitation with anti-PKM2. Immunoprecipitates were probed for phosphorylation at Y105 by immunoblotting with the phospho-specific antibody. (E) FGF inhibitors also reduce phosphorylation of LDHA at the Y10 residue. Phosphorylation of FRS2, an FGFR-interacting protein was included as a control. SDHA served as loading control. (F) Rate of lactate production from explants treated with DMSO (n = 5) or FGF inhibitors PD173074 (5 mM) (n = 6), PD173074 (20 mM) (n = 5), TKI258 (n = 6). (G) Steady-state ATP levels per retina in explants after culture with TKI258 or DMSO. The retinae were transferred to Krebs’-Ringer's with NaN3 or NaCl (untreated group) for 30 min followed by harvest for ATP extraction. n = 7, DMSO+NaCl; n = 9, TKI258+NaCl; n = 9, DMSO+NaN3; n = 9, TKI258+NaN3. Data are Mean±SD. Statistics, Two-way ANOVA with Tukey’s correction. (H) NADPH steady state levels in explants as a percentage of those measured in freshly isolated retina. Explants were treated with DMSO, oxamate, PD173074, TKI258 or left untreated in culture medium. n = 4 groups. Unpaired t-test with Kolmogorov-Smirnov correction for indicated pairs. (I) NADP steady-state levels in explants as a percentage of those measured in freshly isolated retina. Explants were treated with DMSO, oxamate, PD173074, TKI258 or left untreated in culture medium. Oxamate, n = 5; rest, n = 6 groups. Unpaired t-test with Kolmogorov-Smirnov correction for indicated pairs. (J) Blocking glycolysis or FGF signaling reduced EU incorporation in nascent RNA. Explants were treated with DMSO, oxamate, TKI258 or Actinomycin D (RNA Pol II inhibitor) followed by incubation with EU. (K) Quantitative PCR analysis of transcripts to ascertain relative expression of FGF or non-FGF targets (Arr3, Rs1) in explants cultured with or without RPE/Sclera complex (+RPE or –RPE respectively). (L) Ability to produce lactate from neural retina increased when cultured in the presence of RPE/Sclera complex (+RPE) (n = 11) as compared to those that were cultured without the complex (-RPE) (n = 9). Addition of FGF2 in –RPE cultures restored the ability (-RPE+FGF2) (n = 8). Retinal explants were cultured with RPE attached in the explant culture medium. Before transferring them to Krebs’s-Ringer's for lactate estimation, the RPE/Sclera complex was removed and intact neural retina was used. For –RPE conditions, neural retina was cultured in explant medium followed by transfer to Krebs’-Ringer's. FGF2 was added to the explant culture medium but was absent in the Krebs’-Ringer's for -RPE+FGF2 condition. Data depict median in 1–99 percentile box and whiskers plot. Hinges extend between 25th to 75th percentiles. Statistics, Ordinary one-way ANOVA with Tukey’s correction. ONL, outer nuclear layer. INL, inner nuclear layer. GCL, Ganglion cell layer.

https://doi.org/10.7554/eLife.25946.020

To determine if FGF signaling might regulate lactate production, freshly explanted retinae were cultured with TKI258 or PD173074, and lactate secretion was measured. Significantly reduced lactate secretion (Figure 4F) was seen to result from inclusion of either drug. In addition, inhibition of the FGF pathway resulted in increased mitochondrial dependence on ATP steady state maintenance (Figure 4G). Thus, one role for FGF signaling in the adult retina is to promote glycolytic reliance. FGF signaling is required for the maintenance of adult photoreceptors in mice and zebrafish (Campochiaro et al., 1996; Hochmann et al., 2012; Qin et al., 2011). Since it is possible that some of the effects are via regulation of aerobic glycolysis, we examined whether aerobic glycolysis promotes anabolism in the retina. Inhibition of aerobic glycolysis by oxamate treatment or by FGF inhibition resulted in significantly lower steady state NADPH levels- a key cofactor in biosynthetic pathways for lipids, antioxidant responses, and the visual cycle (Figure 4H). We also observed that interference with aerobic glycolysis did not result in an equivalent reduction in NADP+ steady state levels (Figure 4I). The lowering of NADPH level could be attributable to attenuation of the PPP-shunt as a result of decreased glycolytic flux and/or increased usage of NADPH to quench the reactive oxygen species- an unavoidable consequence of increased mitochondrial dependence. We also assessed other effects on cellular anabolism. Nucleotide availability for nascent RNA synthesis was visualized using ethynyl uridine (EU) incorporation after treatment with oxamate and TKI258. Marked reduction in nascent RNA synthesis was evident following inhibition of LDH or FGF signaling (Figure 4I).

Among the large family of FGFs, basic FGF (bFGF/FGF2) has been the most studied in the adult retina. In adult mice and primates, FGF2 is localized to a matrix surrounding photoreceptors and/or is found on their OS (Gao and Hollyfield, 1992, 1995; Hageman et al., 1991). The RPE might contribute to a high FGF2 concentration near photoreceptors via biosynthesis, and/or create a barrier to its diffusion from a retinal source. We first examined the role of the RPE in FGF-signaling. Adult retinal explants were cultured with the RPE/choroid/sclera complex, and expression of FGF target genes in the neural retina was compared with that of explants cultured without the attached complex. In the absence of this complex, the transcripts of known FGF signaling targets displayed reduced steady state levels (Figure 4J). To assess if the reduction in FGF targets was part of a general transcription downregulation or specifically due to dampened FGF signaling, we examined expression of retinoschisin (RS1) or cone arrestin (Arr3), genes expressed at moderate levels in the retina (Blackshaw et al., 2001) (Figure 4J). These genes were not downregulated in the absence of the RPE complex.

The effect of the RPE complex on aerobic glycolysis was analyzed by quantifying lactate production (Figure 4K). Culturing retinae in the presence of the RPE complex resulted in a small, but significant, increase in the ability to produce lactate. Addition of FGF2 in the culture medium was sufficient to increase lactate production from explants cultured without the RPE. Together these data suggest that the RPE/choroid/sclera complex contributes to FGF signaling in the neural retina and that this signaling pathway plays a role in regulating the Warburg effect.

Discussion

Several reports suggest that aerobic glycolysis is a feature of some normal proliferating somatic cells (Agathocleous et al., 2012; Brand and Hermfisse, 1997; Wang et al., 2014b; Zheng et al., 2016), and not just of cancer cells. Our work expands the cell types where aerobic glycolysis can occur to include a mature cell type, the differentiated photoreceptor cell. Like proliferating cells, rod photoreceptors utilize aerobic glycolysis to meet their anabolic needs. A critical aspect of aerobic glycolysis is its ability to be regulated. The data presented here suggest that allostery and FGF signaling are the regulatory mechanisms in the retina. We favor a model where aerobic glycolysis appears to be relevant to photoreceptors not only for organelle maintenance, but likely also helps photoreceptors meet their multiple metabolic demands (Figure 4—figure supplement 1).

In light of this model, it is important to assess the genetic tools that we employed to probe this pathway. Since we drove shRNA expression for Ldha and Pkm2 knockdown from a constitutively active promoter (U6), we speculate that there could be an effect during retinogenesis. We reproducibly observed retinal thinning, indicated by a reduced number of nuclear rows (Figure 1H), especially in very well electroporated retinae. The thinning could be due to perturbation in the cell cycle of retinal progenitor cells, increased cell death, or both. As it is known that there is a role for LDHA and PKM2 in cell proliferation, it is quite likely that such an effect occurred here. The reduced retinal thickness was also apparent in some retinae from dark-reared animals that received the shRNA-encoding constructs against Ldha (Figure 1J). Many photoreceptors that received knockdown constructs against Ldha or Pkm2 showed a significant increase in their OS length after dark rearing. This result argues for a physiological effect due to light exposure having an effect on the OS length, rather than a developmental defect. In addition, our experiments with Tigar gain-of-function, where expression is achieved in a spatiotemporal manner in order to have a minimal effect on retinal development, suggest that the effects of glycolytic perturbation on photoreceptor OSs can be parsed from the confounding effects on retinogenesis.

Aerobic glycolysis in the retina may have implications for blinding disorders. Studies on retinal degenerative disorders indicate that there are metabolic underpinnings to photoreceptor dysfunction, especially those centering around glucose uptake and metabolism (Aït-Ali et al., 2015; Punzo et al., 2009). Furthermore, reducing metabolic stress prolongs survival and improves the function of photoreceptors (Venkatesh et al., 2015; Xiong et al., 2015). In such treated retinae, there is a trend toward upregulation of glycolytic genes (Venkatesh et al., 2015) or metabolites (Zhang et al., 2016). However, a direct cause-and-effect relationship between cell survival and glycolysis has not been established. Our results highlight the metabolic strategies employed by healthy photoreceptors and provide a rational basis for the identification of candidate factors that would further clarify the role of glycolysis in retinal degeneration.

Materials and methods

Plasmids, viruses, in vivo electroporation and transfection

Request a detailed protocol

The synthetic promoter, CAG, consisting of cytomegalovirus (CMV) enhancer, chicken β-actin and rabbit β-globin gene splice acceptor was used for expression and genetic complementation. The expression pattern from this promoter when delivered by electroporation has been described previously (Matsuda and Cepko, 2007). Co-electroporation of a plasmid encoding myristoylated/membrane green fluorescent protein (mGFP) allowed visualization of cells that received the plasmid and marked the inner and outer segments. Co-electroporation rate of plasmids to the retina is close to 100% (Matsuda and Cepko, 2007). Full-length rat Ldhb (rLdhb), human TIGAR, mouse PFKFB3 and human Pkm2 cDNA were obtained from Open Biosystems/GE Dharmacon. Subcloning, epitope tagging and site-directed mutagenesis were carried out by routine molecular biology procedures. For short hairpin (sh) design targeting PKM, and Ldha following resources/software were used: The RNAi consortium, CSHL RNAi central, iRNAi, Invitrogen Block-iT RNAi designer. Designed sh oligos were subcloned in pLKO.1 TRC backbone to be driven by the U6 promoter (Addgene, Cambridge, MA, #10878) and the sequences used in this manuscript are listed in Supplementary file 2. Four hairpin constructs were screened for Ldha and 72 were screened for PKM1/PKM2. Those hairpins that targeted specific mouse sequences but did not target human Pkm2 were chosen. The murine FLAG-tagged PKM1 and PKM2 cDNAs were obtained from Addgene (#44240 and #42512) and subcloned in pCAG-EN. The pyruvate kinase activity from these ORFs has been already reported (Anastasiou et al., 2011).

The plasmids were mixed in equal molar ratios by accounting for their lengths and subjected to Phenol:Chloroform extraction followed by ethanol precipitation and resuspended to a final concentration of 1 mg/mL in Phosphate Buffered Saline. Subretinal in vivo injections and electroporation were carried out as described earlier (Wang et al., 2014a). When possible, the control and experimental constructs were injected in the pups of the same litter and the tail termini were snipped (or left uncut) to identify them later. For knockdown assays or testing expression from plasmids, transfection in HEK293T cells was carried out as using polyethylenimine (PEI). These cells were maintained as a lab stock and were subjected to periodic in-house testing for mycoplasma. Since, these cells were used for protein overexpression and knockdown analyses, concerns of misidentification do not apply to the current work and hence were not checked by third-party testing services.

For making the AAV-mGFP and AAV-TIGAR constructs, the CMV promoter in the empty AAV-MCS8 vector (Harvard Medical School DF/HCC DNA Resource Core) was replaced with the bovine rhodopsin promoter (Matsuda and Cepko, 2007). Woodchuck hepatitis virus posttranscriptional response element (WPRE) was added to enhance expression. Capsid type 8 AAVs were produced and titered as described previously (Xiong et al., 2015). For subretinal injections of AAV, ~3.5 × 106–5 × 106 particles (based on genome copies) per eye were used. P6 pups were injected in order to transduce cells after the proliferative phase of retinogenesis so as to minimize any detrimental effects on cell division and dilution of replication-incompetent viruses. The extent of infection was assessed with a Keeler indirect ophthalmoscope using the cobalt blue filter and Volk 78 diopter lens on non-anesthetisized animals. Mice with edge-to-edge infection were tagged and used subsequently for lactate assays and immunoblotting.

Mice and animal husbandry

Request a detailed protocol

Timed pregnant, wild-type CD1 female mice were obtained from Charles River Laboratories, Boston, MA, and P0-P1 pups thereof were used in electroporations. C57BL/6J and the two-color Cre reporter mouse Gt(ROSA)26Sortm4(ActB-tdTomato,-EGFP)Luo/J (referred to as mT/mG and described previously [Muzumdar et al., 2007]) were obtained from the Jackson Laboratories (JAX), Bar Harbor, ME. Ldhafl/fl (Wang et al., 2014b), Pkm2fl/fl (Israelsen et al., 2013), Rod-cre (Le et al., 2006) mice have been described before. Rod-Cre; Ldhafl/fl and Rod-cre; Pkm2fl/fl mouse lines were established. For experimentation, these mice were backcrossed with Ldhafl/fl or Pkm2fl/fl parents and Cre+ and Cre F1 progeny were used to ensure equivalent allelic copies of the Cre transgene, minimum genetic difference and ease of age-matching by using the siblings. Animals were housed at room temperature with 12 hr light and 12 hr dark cycle. Light inside the cages in the room varied from 0 to 3 lx in the cage farthest from the light source to 300 lx in the cage closest to it. As a practice, the electroporated mice inhabited rack spaces where light intensity in the cages varied from ~175 to~235 lx. At weaning, the mice were segregated according to their sexes, thus a cage usually had the control and electroporated pups from the same litter. Tamoxifen injections were carried out as described previously (Matsuda and Cepko, 2007). For dark rearing, electroporated animals were raised with their mothers until P11, when the eyes started to open. Following this, they were transferred to animal housing maintained in darkness until weaning age, when they were weaned and group housed in dark until indicated times for harvest. In order to minimize effects due to circadian regulation of OS growth, the light and dark-reared animals from all the groups (control, LDHAsh and PKM2sh) were harvested on the same day and within 3 hr of each other. Water and chow were available ad libitum. Animal care was following institutional IACUC guidelines.

Dissections and adult explant cultures

Request a detailed protocol

Wild-type, pigmented C57BL/6J mice (JAX) were used for explant cultures since presence or absence of RPE was easily discernable. For adult retinal cultures, P23-P28 animals were euthanized by CO2 asphyxiation and freshly enucleated eyes were dissected rapidly in Hanks buffered saline solution (HBSS) (Invitrogen, Carlsbad, CA). Extraocular tissue was trimmed off and the cornea and iris were carefully removed. Sclera along with the RPE was gently removed. This was done primarily for two reasons: (1) In our assays the presence of Sclera/RPE complex significantly reduced the efficacy of drug treatments and, (2) secreted lactate was not detected from freshly isolated eyecup with intact sclera. Lens was retained to keep the sphericity of the retina for uniform access to the medium. Explant medium consisted of Neurobasal-A, 0.2% B27 supplement, 0.1% N2 supplement, 0.1% Glutamax and penicillin/streptomycin (all Invitrogen). Retinae were incubated in freshly prepared explant medium constantly supplied with 95% O2 + 5% CO2 (Medical Technical Gases) at 37°C in a roller culture system (B.T.C Engineering, Cambridge, UK) for indicated times. At the end of incubation period, the lens was removed and the retinae were quickly rinsed with prewarmed Krebs’ Ringers medium (98.5 mM NaCl, 4.9 mM KCl, 2.6 mM CaCl2, 1.2 mM MgSO4, 1.2 mM KH2PO4, 26 mM NaHCO3, 20 mM HEPES, 5 mM Dextrose) saturated with 95% O2. Retinae were again incubated in 0.5 mL Krebs’ Ringers medium for 30 min in roller culture with 95% O2 supplied. The supernatant and retinae were rapidly frozen separately at the end of the experiment. DMSO was used as vehicle control for water-insoluble solutes. Sodium oxamate or sodium azide was dissolved in the medium. Equimolar amount of sodium chloride was used as control for osmotic pressure, a colligative property. For +RPE experiments, the extraocular tissue was trimmed off, cornea and iris removed and the eyecups were incubated in the explant culture medium. At the end of the incubation, the RPE/sclera complex was removed along with the lens and the neural retina was incubated in the oxygenated Kebs’s Ringers medium for 30 min as described earlier to assay secreted lactate. Thus, our experiments assess the effect of RPE/sclera complex on the ability to produce lactate by neural retina.

Drugs

Sodium Azide (20 mM, Sigma-Aldrich), Sodium Oxamate (50 mM, Sigma-Aldrich), FX11 (10 μM, Calbiochem, San Diego, CA), BMS 536924 (5 μM, Tocris, Minneapolis, MN), Afatinib (5 μM, Selleckchem, Houston, TX), Dovitinib/TKI258 (5 μM, Selleckchem), Dasatinib (5 μM, Selleckchem), PD173074 (5 μM or 20 μM, Selleckchem), Actinomycin D (5 μM, Sigma-Aldrich, St. Louis, MO), FGF2 (2 μg/mL, Cell Signaling, Danvers, MA).

Immunoprecipitation and immunoblotting

Request a detailed protocol

BL/6J retinae without RPE were homogenized in Lysis buffer (5 mM HEPES, 1 mM DTT, 1 mM ATP, 5 mM MgCl2, 1% glycerol, Complete Protease Inhibitor (Roche) and PhosStop phosphatase inhibitor (Roche, Basel, Switzerland). Immunoprecipitation was carried out using rabbit anti-PKM2 and rabbit IgG isotype control followed by sheep anti-rabbit-conjugated Dynabeads (Life Technologies). Immunoprecipitates were boiled and loaded on 10% SDS-PAGE gels followed by transfer on Hybond nitrocellulose membranes (GE Amersham, Amersham, UK). Membranes were blocked with 5% non-fat milk in 1X Tris Buffered Saline +0.1% Tween-20. A conformation-specific mouse-anti rabbit secondary and HRP-conjugated goat-anti-mouse (Jackson Immunoresrearch, 1:10,000) tertiary antibodies were used followed by Enhanced Chemiluminescent (ECL) detection using substrate from GE Amersham.

Immunohistochemistry

Request a detailed protocol

Enucleated eyes were fixed overnight at 4°C in 4% formaldehyde. The eyes were passed through an increasing concentration of sucrose (5%, 15%, 30%) followed by equilibration in 1:1 30% sucrose: OCT (Sakura Finetek, Torrance, CA) and frozen on dry ice. Eighteen micron cryosections were cut using a Leica CM3050S cryostat. Antibodies used are listed in Supplementary file 3. Heat-mediated antigen retrieval at pH 8 was carried out. For HRP staining, Cell and Tissue staining kit (R and D systems) was used. Confocal images were acquired on Zeiss LSM710 or LSM780 inverted microscope. The intensity and pixel saturation were calibrated for inner and outer segment label (mGFP) so that details in these cellular features were retained. Thus, due to intense signal of mGFP in the outer segments, the labeling in other cells of the inner retina seems variable and less bright despite electroporation known to target these cells. Images were processed on ImageJ. Maximum intensity projections are depicted. Colocalization was confirmed by individual merges of coplanar sections along the z-axis. For IS/OS length measurements, the orthogonal projections of sections were used. The projections spanning the entire IS/OS volume ensure changes due sectioning angle have a minimal effect. Multiple quantifications across the electroporated field were done for at least three retinae. Expression by IHC was confirmed in both CD1 (albino) and BL/6J (pigmented) mice. Sclera and RPE were preserved in electroporated eyes to ensure that outer segments were not ripped during the dissections. For all procedures involving antibodies, multiple antibodies were sourced and tested whenever possible (Supplementary file 3). Previously published antibodies were included and cross-verified with other commercially available antibodies. IHC data were always cross-verified with RNA ISH.

In situ hybridization

Request a detailed protocol

In situ hybridization was carried out as described earlier (Blackshaw et al., 2001). Probe sequences are presented in Supplementary file 4. For Pfkfb1, Pfkfb2, Pfkfb4, Srsf3 and Ptbp1, tyramide amplification (Perkin, Waltham, MA) was used. Bright-field images were acquired on Nikon Eclipse E1000 microscope.

ATP, Lactate and NADPH assay

Request a detailed protocol

For ATP estimation, individual retinae were rapidly frozen in liquid nitrogen at the end of the assay. ATP was measured using ATP bioluminescence kit CLS II (Roche/Sigma-Aldrich). For secreted lactate estimation the retinae were incubated in Krebs’ Ringers medium after indicated treatments. The supernatant from above was used with Lactate assay kit (Eton Bioscience, San Diego, CA). Amount of lactate produced in 30 min was assayed. Intracellular lactate was estimated for AAV-transduced retinae because a large number of mice had to be injected and screened for complete, edge-to-edge infection. Thus, infected retinae at specific age were harvested and frozen as they became available. All these retinae were harvested 4–5 hr after lights were turned on in the facility to minimize variability due to possible cyclical diurnal changes in metabolism. Two to three retinae were pooled into a group and frozen together. Five such groups (n = 5) were used for assaying lactate after AAV-mGFP and AAV-TIGAR infection. The retinae were homogenized with the Lactate Assay buffer (Fluorometric Lactate Assay kit, abcam, Cambridge, MA). A small aliquot was removed for protein estimation and subsequent immunoblotting and the remainder was passed through 10 kDa protein filtration column (abcam) to remove proteins and thus minimize interference due to endogenous lactate dehydrogenase in the lactate assay. For protein estimation, Qubit protein assay (Invitrogen) was used since it is not affected by the presence of detergents in the Lactate Assay buffer. NADP and NADPH was assayed using Fluoro NADP/NADPH kit (Cell Technology, Fremont, CA) following manufacturer’s instructions. The quantifications for NADP and NADPH were made separately and thus represent different retinae and treatments.

COX and SDH histochemistry

Request a detailed protocol

Histochemistry on fresh and unfixed retinal tissue was carried as described earlier for brain tissue (Ross et al., 2010). The assay relies on the ability of functional cytochrome oxidase to catalyze oxidative polymerization of 3,3'-diaminobenzidine (DAB) (an electron donor) to brown indamine product. Succinate dehydrogenase assay is based on the ability of this enzyme to oxidize supplied succinate and in turn reduce a ditetrazole (NBT) to dark blue diformazan using phenazine methosulfate (PMS), an intermediate electron carrier.

5-Ethynyl uridine (EU) labeling

Request a detailed protocol

Explants were cultured with indicated drug or DMSO for 5 hr followed by 1 mM EU (Life Technologies) with the drug for additional 2.5 hr. The retinae were fixed, cryosectioned and processed for label detection using Click chemistry reagents (Life Technologies).

Quantitative RTPCR

Request a detailed protocol

RNA was isolated using TRIzol reagent (Life Technologies, Carlsbad, CA) from 3 to 4 retinae. Two μg RNA was subjected to cDNA synthesis using SuperScript III reverse transcriptase and random hexamers. QPCR was performed using power SYBR Green PCR Master mix (Applied Biosystems, Foster City, CA) on a 7500 Fast Real-Time PCR System (Applied Biosystems). Primer sequences are provided in Supplementary file 5. Rpl13a was used as internal reference and freshly isolated retinal tissue was used as calibrator sample. Expression ratio was calculated using 2-ΔΔCt method. For each target gene, three technical replicates were simultaneously assayed to arrive at the average value for a biological replicate. Mean of three biological replicates was used to derive the Ct value of each target.

Rod isolation and cDNA synthesis

Request a detailed protocol

P0 CD1 mice were electroporated with Rho-dsRed plasmid which encodes for dsRed, driven by bovine rhodopsin promoter, which results in retinas with patches of dsRed expression only in rod photoreceptors (Matsuda and Cepko, 2004). Once they reached adulthood, mice were then euthanasized via CO2 asphyxiation and the retinas were rapidly removed. The retinas were incubated for 5 min at 37°C in Hank’s Balanced Salt Solution (HBSS) supplemented with 10 mM HEPES and 5 mM EDTA and then gently triturated with a P1000. The dissociated retina was allowed to settle on sylgard-coated petri dishes. Rods expressing the dsRed reporter were identified by their red fluorescence using an inverted microscope and hand-pipeted directly into lysis buffer, and their cDNA amplified using the previously described protocol that utilizes oligo dT priming (Goetz and Trimarchi, 2012).

Data collection and statistics

Request a detailed protocol

Data collection was from non-randomized experiments. The primary experimenters were not blinded to treatments. No statistical methods to predetermine sample size were employed. No assumptions for potential outliers were made and hence all data points were included in analyses and depicted. Normality of data distribution was tested using D’Agostino-Pearson omnibus test. Non-parametric statistics were used when Gaussian distribution of data points could not be obtained. p-value denoted as: Not significant (NS), p>0.05; *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
    Isozyme patterns of normal, benign, and malignant human breast tissues
    1. D Balinsky
    2. CE Platz
    3. JW Lewis
    (1983)
    Cancer Research 43:5895–5901.
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
    Retinal degeneration in transgenic mice with photoreceptor-specific expression of a dominant-negative fibroblast growth factor receptor
    1. PA Campochiaro
    2. M Chang
    3. M Ohsato
    4. SA Vinores
    5. Z Nie
    6. L Hjelmeland
    7. A Mansukhani
    8. C Basilico
    9. DJ Zack
    (1996)
    Journal of Neuroscience 16:1679–1688.
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
    LACTIC DEHYDROGENASES: FUNCTIONS OF THE TWO TYPES RATES OF SYNTHESIS OF THE TWO MAJOR FORMS CAN BE CORRELATED WITH METABOLIC DIFFERENTIATION
    1. DM Dawson
    2. TL Goodfriend
    3. NO Kaplan
    (1964)
    Science 143:929–933.
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    Single-cell profiling of developing and mature retinal neurons
    1. JJ Goetz
    2. JM Trimarchi
    (2012)
    Journal of Visualized Experiments, 10.3791/3824, 22546911.
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
    Mouse opsin promoter-directed cre recombinase expression in transgenic mice
    1. YZ Le
    2. L Zheng
    3. W Zheng
    4. JD Ash
    5. MP Agbaga
    6. M Zhu
    7. RE Anderson
    (2006)
    Molecular Vision 12:389–398.
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
    Lactate released by müller glial cells is metabolized by photoreceptors from mammalian retina
    1. CL Poitry-Yamate
    2. S Poitry
    3. M Tsacopoulos
    (1995)
    Journal of Neuroscience 15:5179–5191.
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
    The cellular and compartmental profile of mouse retinal glycolysis, tricarboxylic acid cycle, oxidative phosphorylation, and ~P transferring kinases
    1. EM Rueda
    2. JE Johnson
    3. A Giddabasappa
    4. A Swaroop
    5. MJ Brooks
    6. I Sigel
    7. SY Chaney
    8. DA Fox
    (2016)
    Molecular Vision 22:847–885.
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67

Decision letter

  1. Jeremy Nathans
    Reviewing Editor; Johns Hopkins University School of Medicine, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Glycolytic reliance promotes anabolism in photoreceptors" for consideration by eLife. Your article has been favorably evaluated by Richard Aldrich (Senior Editor) and three reviewers, one of whom, Jeremy Nathans, is a member of our Board of Reviewing Editors. One of the reviewers (Reviewer 1) has agreed to reveal his identity: James B. Hurley.

As you will see, all of the reviewers were impressed with the importance and novelty of your work. I am including the three reviews (lightly edited) at the end of this letter, as there are a variety of specific and useful suggestions in them. Some of the comments can be handled with changes in the text or responses to the reviewers but others require additional experimental work. It may help us to assess your prospects for a timely revision if you write back with responses and a plan to address the concerns of the reviewers.

Reviewer #1:

This outstanding, highly significant, timely and well written report aims to connect glycolytic metabolism in retinal photoreceptors with their size. The investigators link recent findings in the field of cancer metabolism to the influence of metabolism on growth and survival of primary sensory neurons. They use sophisticated and appropriate genetic approaches to perturb the normal metabolic functions of photoreceptors in vivo. They show that deficiencies in aerobic glycolysis alter photoreceptor growth. This is a major advance for understanding what is needed for homeostasis in primary sensory neurons. The authors should address the following specific comments.

1) Introduction, second paragraph: Please note that Wang et al., (1997) PMID: 9179314 showed that aerobic glycolysis in the retina originates in the outer retina.

2) For the experiments in Figure 1 please specify concentrations of oxamate and azide.

3) This statement is confusing: "A tetramer of all ldha subunits has high affinity for pyruvate and a higher Vmax for pyruvate reduction to lactate." Is this referring to a higher Vmax than ldhb or a higher Vmax for lactate to pyruvate? Also, it isn't clear from this which one has a higher affinity for pyruvate, Ldha or Ldhb tetramers? Please spell out the kinetic differences between the LDH isoforms in more detail.

4) The legend for supplement Figure 1—figure supplement 2D should state the conclusion of that experiment – that the Cre is not expressed in cones.

5) I'd think it would be helpful for the authors to compare more comprehensively the top two rows of panel H of Figure 1. It looks like the overall thickness of the retina is smaller in the second row. Is that real and reproducible?

6) How do the authors correct for variations in the angle of sectioning when measuring the lengths of the OS? It would be useful for at least a few of the experiments to present the OS and IS length as a% of the total thickness of the retina. Also, are the morphologies of the Muller cell or RPE affected?

7) For the electroporation of shRNA experiments for LDHA the authors highlight how electroporation occurs in only 15-30% of the retina. That means there must be some regions where there is a transition from high to low LDHA expression. Can the authors tell whether or not the loss of LDHA is cell autonomous or is the length of the outer segment influenced by the amount of lactate produced by its neighboring rods? Using anti-rhodopsin antibodies (see next comment) to label the OS in IHC may help reveal the phenotype in these regions.

8) The authors use mGFP expression to measure the OS and IS length. Would it be possible/worthwhile to confirm a couple of the findings by simple staining some of the sections with a rhodopsin antibody that would very clearly label the OS and then quantifying the lengths of the OS to make sure that type of measurement is consistent with the measurements made from the expressed mGFP? Please either address this experimentally or explain why it would not be practical or necessary.

9) For the experiments in Figure 1J,K and also other experiments where outer segment and inner segment length were measured – were all the retinas collected at the same time of day?

10) Maybe I overlooked it, but I could not find the part of the text that refers to Figure 2—figure supplement 1 panels C-F.

11) Subsection “Nonequivalent roles of pyruvate kinase isoforms”, first paragraph: Please also cite recent paper from the Rajala lab (PMC5121888) that show expression patterns of PKM1 and PKM2 and tyrosine phosphorylation of PKM2. Also please cite the relevant IOVS paper by Casson et al. PMID: 26780311. Another paper that should be cited regarding PKM2 in the retina is Molecular Vision paper by Rueda et al. (PMID: 27499608)

12) The difference in the overall thickness of the retina is particularly obvious in Figure 3—figure supplement 1G (the rescue with huPKM2). It seems like the whole retina is affected by the PKM2 expression. Is that correct? The authors should address the possibility that the whole retina is affected in a non-cell autonomous way. (Minor note: the legend refers to the boom panel on the right but it is on the left).

13) Subsection “Nonequivalent roles of pyruvate kinase isoforms”, fourth paragraph: PKM2fl/fl and Rod-cre mice were used to knockout PKM2 in rods for the lactate measurements. Please note whether the rods were shorter in these experiments. Also, if I'm right that the overall thickness of the retina changes do that also occur in the PKM2fl/fl/rod-cre retinas?

14) In the subsection “Fibroblast growth factor signaling regulates anabolism”, the authors should cite a recent paper, PMC5121888, that also confirms that Y105 of PKM2 is phosphorylated in mouse retinas.

15) Please include MW marker positions on Figure 4B. Also, please explain what the "isotype-matched" means.

16) The authors should be careful to not over-interpret data in some of their descriptions. For example, the authors state "Impact of glycolytic perturbation on nucleotide availability was directly visualized[…]". Since FGF signaling can influence other processes besides glycolysis I think this is over-interpreting the data. It would be better to just say that FGF receptor signaling influences nucleotide availability and this could be linked to limitation of glycolysis. Alternatively the connection could be strengthened by evaluating the effects of PKM2 or LDHA inhibition on EU incorporation.

Reviewer #2:

Vertebrate photoreceptors are among the most metabolically active cells, exhibiting a high rate of ATP consumption. This is coupled with a high anabolic demand, necessitated by the diurnal turnover of a specialized membrane-rich organelle, the outer segment, which is the primary site of phototransduction. It is not clear to date that how photoreceptors balance their catabolic and anabolic demands. The current study has attempted to address this fundamental issue in photoreceptor biology. The authors have used several biochemical, immunological, genetic and viral transduction methods to address the importance of glycolysis on the outer segment biogenesis. The study is interesting but failed to support the authors claim that glycolysis regulates the outer segment biogenesis.

1) In these studies, authors have examined the isoform expression of LDH, PKM2, then phosphorylation state of PKM2 under dark- and light-adapted conditions, identified FGF signaling promotes PKM2 phosphorylation and splicing regulation of PKM1 and PKM2. Some of these studies have been done before by other labs (see below), which the authors did not acknowledge or reference in their manuscript. These include:

• Identification of LDH isoforms (Casson RJ et al. 2016 IOVS).

• PKM1 and PKM2 isoform characterization, light-dependent tyrosine phosphorylation of PKM2 (Rajala et al. 2016 Sci Rep).

• The authors claimed in this study that fibroblast growth factor (FGF) signaling was found to regulate glycolysis through phosphorylation of PKM2. This finding is not novel as it has been elegantly shown in tumor cells that FGFR regulates PKM2 phosphorylation (Hitosugi et al. 2009 Sci Signal).

• Regulation of PKM1 and PKM2 splicing (Su et al. (2017 Mol Cell Biol)

The authors are urged to cite these earlier references and give proper credit for these studies. They could discuss how the published results are similar or differ from their observations in this manuscript.

2) LDH isoforms identification has previously been reported (Casson RJ et al. 2016 IOVS). The authors must cite this manuscript.

3) The authors stated in the manuscript (subsection “Lactate producing isoform of Ldh in photoreceptors”, last paragraph) the recombination efficiency with rod-cre varied between 50-90%, but that is not correct. The rod-cre used in these studies will not recombine more than 50%. The authors have shown only protein expression by Western blots. They need to show the deletion by immunohistochemistry.

4) The authors have shown that lactate production was significantly reduced in conditional LDH-A mice. Why did the authors not study OS biogenesis in LDH-A deleted mice? The shRNA strategy is not well justified. Have the authors examined the OS in LDH-A KO mice? Generally, the shRNA approach may not knock-down completely the gene of interest but conditional deletion will? The authors observed the shortening of OS. Could this be an off-target effect? LDH-A is also expressed in other layers of the retina (INL and IPL).

Immunohistochemistry is not the ideal way to demonstrate OS length; hence the authors measured IS+OS. The authors should use ultrastructural studies, such as EM or high resolution LM to demonstrate the OS length phenotype. Some of the micrographs show thinning of the outer nuclear layer thickness (e.g., DAPI stained sections in Figure 1H, 1J, 2C, and 3C, suggesting retinal degeneration in these genetically modified retinas, which could argue against shortening of OS length. Did the authors do TUNEL or any other test for dying photoreceptor cells?

5) Could shRNA knock down in other retinal layers (may be Muller cells) may indirectly affect the structure of OS?

6) Figure 1K – There is no evidence of disc shedding in this experiment. Could there be less opsin trafficking to the OS? Such a possibility cannot be ruled out. This experiment is overstated.

7) – The statement "lactate production by the photoreceptors cannot be attributed to lack of mitochondrial activity." The authors have done experiments on (Figure 1L) wild- type retinas. Have they carried out these experiments in LDH-A-knockdown or KO mice?

8) Figure 2A and F – what is the rationale to regulate TIGAR expression spatially and temporarily? In Figure 2F the authors used AAV-mediated expression of TIGAR. Lactate levels were done in AAV-TIGER (Figure 2G) but not for Figure 2A? It is very confusing, and there was no rationale provided for these experiments. It seems that authors may have difficulty in measuring lactate levels for the inducible expression system?

9) Figure 3A – PKM1 and PKM2 expression has recently been reported (Rajala 2016 Sci Rep). The authors have not cited this reference.

10) The authors have shown the developmental expression of PKM2 and PKM1 on western blots (Figure 3B), which is not the ideal way to show the developmental expression. If the authors wanted to show this, they should provide immunohistochemistry or ISH.

11) The authors used rod-cre to delete PKM2 and measured LDH activity (Figure 3G). For structural studies, they used shRNA and examined OS length (Figure 3G). These studies are not convincing. Why did they not observe similar shRNA effects with conditional PKM2 KO mice?

12) There is no indication of how much PKM2 is deleted or knocked down. In the absence of these experiments, it is very difficult to interpret the data. Moreover, the authors did not carry out any functional studies, such as ERG to examine the role of PKM2 in photoreceptor functions?

13) Figure 3—figure supplement 2 does not add any new information. The authors' data show opposite expression of these splicing factors. There was a study recently published showing that RBM4 Regulates Neuronal Differentiation of Mesenchymal Stem Cells by Modulating Alternative Splicing of Pyruvate Kinase M (Mol Cell Biol 2017).

14) The authors stated that PKM2 deletion upregulates PKM1, but has no effect on photoreceptor structure (Figure 3—figure supplement 1H). On the other hand, forceful expression of PKM1 had a reduction in the length of OS? How do authors explain this discrepancy?

15) Figure 4 – FGF signaling – Authors have identified that FGF signaling promotes the phosphorylation of PKM2. It is not a novel finding. It has been shown in tumor cells (Hitosugi et al., 2009). The authors have not acknowledged this information in the current manuscript.

16) Figure 4C – PKM2 undergoes a light-dependent tyrosine phosphorylation on Tyr105. These studies have recently been reported by Rajala et al. 2016 (Sci Rep). The authors have not acknowledged this study.

Reviewer #3:

This manuscript provides novel and interesting data on the reliance of aerobic glycolysis for photoreceptor outer segment renewal. Overall, the paper is very good and a significant contribution. However, there are some significant problems that need addressing before the results and conclusions that are presented can be accepted. In addition, there are several additional items that are off putting and overstepping the presentation and results.

First, the authors are only addressing aerobic glycolysis in the rod inner segments and outer segments. There does not appear to be any data for cones or rod and cone photoreceptor synapses. So, the title should reflect this and state something akin to "Aerobic glycolytic reliance promotes anabolism in rod photoreceptor outer and inner segments."

Second, the authors should provide more detailed and important information in the Methods. This information is critical for proper interpretation of the results. For example, what the light level in the animal room and cages, what time of day were the mice sacrificed, what area and quadrant of the retina was quantified for OS-IS measurements, how was this location established, what are the measurements for OS and IS alone as they could both change, how many non-adjacent sections per retina were quantified per mouse, what was the volume of incubation for the lactate assay, what type of confocal was used and how many Z-stacks were there per image, what was the exact age of mice used for the histology as in the subsection “Dissections and adult explant cultures”, the authors stated that they used P23-P28 mice. According to LaVail 1973 J Cell Biol, the outer segments in the C57BL/6J mice reach their rate of synthesis and disposal at around P21-25.

Third, the authors conclude from looking at their retinas 42-45 days following electroporation that the shortened rods were due to decreased synthesis due to decreased aerobic glycolysis. This cannot be accepted at face value without conducting two additional experiments: show rod disc synthesis results [e.g., per Young and Bok studies] and rod phagocytosis results [e.g., per LaVail studies].

Fourth, please explain why eye opening occurred on day 11 in these studies. Most published mouse paper find that this event occurs at postnatal day 14 +/- 1 day.

Fifth, the choice of words in several places throughout the manuscript are overstated or not properly used. This often results from lack of citation of previously conducted work. For example, they state that "The cell types that carry out aerobic glycolysis in the normal adult retina have not been determined." This is just false. The work of Rueda et al. published in Mol Vis in 2016 that shows cell-type glycolysis, aerobic glycolysis, high energy transferring kinases and oxidative phosphorylation in over 20 different compartments and cells in the retina. A summary figure clearly shows all of this data. However, the authors have not cited this manuscript anywhere in their manuscript.

For example, they state that "surprisingly the steady state levels of ATP…did not differ from controls." Winkler in J Gen Physiol 1981 showed that retinal ATP levels are steady under most conditions: not surprisingly since Lowry and co-workers demonstrated in the 1960s that this energy measure is the last to change even after anoxia.

For example, in the subsection “Aerobic glycolysis in the retina”, the authors incorrectly stated that adenylate kinase (AK) synthesizes ATP. The do not acknowledge that mouse retina also expresses creatine kinase (CK). The authors need to state that CK and AK serve to regenerate the ATP by reversible reactions that respond to the law of mass action. The implication of these enzyme reactions need to be explained as these enzymes may respond under conditions of high ATP hydrolysis or raising [ADP] (E.g., see the work of Wallimann, Linton et al. 2010; Rueda et al., 2016).

For example, the authors state that "presumably as a result of less shedding". This results from the process of "photostasis' as published in Exp Eye Res 1n 1986 by Penn and Williams.

For example, the authors state that "Our work expands the cell types where aerobic glycolysis can occur to include a mature cell type, the differentiated photoreceptor cell". As noted, the authors did not cite the prior work of Rueda et al. published in Mol Vis in 2016. This work also pyruvate competition in LDH histochemistry assays and confocal to demonstrate that Ldh-A was preferentially located in photoreceptors…not cited in the first paragraph of the subsection “Lactate producing isoform of Ldh in photoreceptors”, as confirmatory. So the current manuscript "confirms and definitely expands" would be the correct terminology.

For example, the Abstract says that the process of photoreceptor catabolism and anabolism is "poorly understood". It should be more precise as there are many studies on these process, but not in the context of rod outer segment biosynthesis.

Sixth, no retinal references are provided for COX and SDH histochemistry. Several previous papers conducted these experiments in rodents (SDH in rat: Hansson, Exp Eye Res 1970: COX in rat: Chen et al. Acta Ophthalmol 1989; COX in mouse: Rueda et al., Mol Vis 2016). The latter paper in adult mouse was not cited in the subsection “Functional mitochondria in photoreceptors”, although quantitative layer-by-layer COX activity was measured and presented. The authors stated that SDH activity was not different between inner and outer retina, but this does not agree with prior published results.

Seventh, the authors state that "PKLR transcripts were not detected in retina (data not shown)". However, they were found in the retina by other investigators (see Figure 1 of Rueda et al. Mol Vis 2016). No comment is made to acknowledge the difference.

Eighth, the authors should have used a higher resolution microscopy to show specifically the changes observed in the outer segments and possible exclude or determine the possible changes in inner segments and inner segment-mitochondria due to loss of aerobic glycolysis.

Ninth, please clarify in the first paragraph of the subsection “Allosteric regulation of glycolysis in photoreceptors”, the three listed criteria points.

Tenth, Ross et al. 2010a should be 2010 as there is no "b".

https://doi.org/10.7554/eLife.25946.029

Author response

Reviewer #1:

This outstanding, highly significant, timely and well written report aims to connect glycolytic metabolism in retinal photoreceptors with their size. The investigators link recent findings in the field of cancer metabolism to the influence of metabolism on growth and survival of primary sensory neurons. They use sophisticated and appropriate genetic approaches to perturb the normal metabolic functions of photoreceptors in vivo. They show that deficiencies in aerobic glycolysis alter photoreceptor growth. This is a major advance for understanding what is needed for homeostasis in primary sensory neurons. The authors should address the following specific comments.

1) Introduction, second paragraph: Please note that Wang et al., (1997) PMID: 9179314 showed that aerobic glycolysis in the retina originates in the outer retina.

The work by Anders Bill and colleagues in the aforementioned publication (Wang et al., 1997, Acta Physiol. Scand., 160, 75-81) explored the production of lactate and consumption of oxygen and glucose in dark and light in the pig retina. They used direct measurement of metabolites (lactate and glucose) and oxygen from the vorticose vein. These vessels also drain from the uvea including anterior uvea. It is known that the lens cells are glycolytic after they expunge much of their organelles. Though light should not affect the glycolytic state of these cells, it is still an assumption until proven otherwise. The other effect that should be considered is that of the effect of constriction of sphincter pupillae muscles following light stimulation. How much lactic acid is produced from these muscles in their contracted state and how much do they contribute to the free lactate in the serum? Even if the contribution towards the lactate pool from the outer retina proves significant over that of the anterior uvea, can one be certain that it is attributable to photoreceptors? One can also interpret these results in the light of the lactate shuttle hypothesis i.e. Mueller glia secrete lactate (the directionality of the flow can be explained by consumption of lactate by the photoreceptors). Given that photoreceptors have an increased energy demand in the darkness, they consume more of the lactate, resulting in less being secreted into the blood stream. This would align with the observation of Wang et al. as they note ~60% decrease in lactate levels in darkness. Lastly the contribution of RPE to the steady state free lactate levels in the blood is not known, but should be considered when retinal physiology models are constructed. We do not contest the validity of their observations. But from our point-of-view, these data do not help to discriminate between two opposing models of photoreceptor physiology based on glycolysis and lactate production or consumption. Since that does not change our conclusion laid out in the second paragraph of the Introduction, that photoreceptors are assumed to rely on aerobic glycolysis and a systematic analysis of relative propensities of photoreceptors and other cell types, especially Mueller glia, for aerobic glycolysis has not been carried out, we had refrained from citing this work in the interest of space and number of articles that should ideally be referenced. Inclusion of the caveats listed above would have taken up quite a bit of space and we did not want to be overly critical of the previous work in this area by pointing these out. Incidentally, these caveats also apply to many such studies wherein the entire eye, or the entire retina, or the retina and associated tissues (e.g. RPE) were assayed.

2) For the experiments in Figure 1 please specify concentrations of oxamate and azide.

The concentrations of these agents and others are reported in Materials and methods (subsection “Drugs”).

3) This statement is confusing: "A tetramer of all ldha subunits has high affinity for pyruvate and a higher Vmax for pyruvate reduction to lactate." Is this referring to a higher Vmax than ldhb or a higher Vmax for lactate to pyruvate? Also, it isn't clear from this which one has a higher affinity for pyruvate, Ldha or Ldhb tetramers? Please spell out the kinetic differences between the LDH isoforms in more detail.

Given that our approach relied on exploiting relative differences between Ldha and Ldhb isoforms, we agree that a clearer explanation is warranted. We have made the suggested changes (subsection “Lactate producing isoform of LDH in photoreceptors”, first paragraph).

4) The legend for supplement Figure 1—figure supplement 2D should state the conclusion of that experiment – that the Cre is not expressed in cones.

We had refrained from stating conclusions in the figure legends, as we prefer that data in a figure show the conclusion. To be consistent, we are avoiding it here as well. We have stated the conclusions in the text body (subsection “Lactate producing isoform of LDH in photoreceptors”, last paragraph).

5) I'd think it would be helpful for the authors to compare more comprehensively the top two rows of panel H of Figure 1. It looks like the overall thickness of the retina is smaller in the second row. Is that real and reproducible?

We agree with the reviewer’s observation that the thickness of the retina is reduced. We have now addressed this issue in the Discussion (second paragraph).

6) How do the authors correct for variations in the angle of sectioning when measuring the lengths of the OS? It would be useful for at least a few of the experiments to present the OS and IS length as a% of the total thickness of the retina. Also, are the morphologies of the Muller cell or RPE affected?

We agree with the Reviewer’s concern that quantification involving OS length should factor in all the sources of variation. Changes in sectioning angle can introduce significant variability especially in the photoreceptors lying at the cut surface (closest to the imaging objective). We have addressed this issue in Materials and methods subsection “Immunohistochemistry”. Confocal imaging of the electroporated photoreceptors and orthogonal projections of photoreceptor volume along the Z-axis, enabled us to sample greater depth and image more photoreceptors in their entirety in the retinal tissue, rather than only those on the cut surface. We sampled multiple photoreceptors in any given field, multiple fields in an electroporated patch, and multiple retinae that were independently sectioned. In addition, we included all data points without setting statistical cutoffs for outliers (Materials and methods subsection “Data collection and statistics”). We also depict all the data points to give readers the idea of the data spread. In summary, we have made efforts to minimize variability in our data collection and include all data points for statistical analysis.

7) For the electroporation of shRNA experiments for LDHA the authors highlight how electroporation occurs in only 15-30% of the retina. That means there must be some regions where there is a transition from high to low LDHA expression. Can the authors tell whether or not the loss of LDHA is cell autonomous or is the length of the outer segment influenced by the amount of lactate produced by its neighboring rods? Using anti-rhodopsin antibodies (see next comment) to label the OS in IHC may help reveal the phenotype in these regions.

This comment and the following one addressed together as 1 response.

8) The authors use mGFP expression to measure the OS and IS length. Would it be possible/worthwhile to confirm a couple of the findings by simple staining some of the sections with a rhodopsin antibody that would very clearly label the OS and then quantifying the lengths of the OS to make sure that type of measurement is consistent with the measurements made from the expressed mGFP? Please either address this experimentally or explain why it would not be practical or necessary.

We have included rhodopsin staining for the LDHA knockdown retina (Figure 1—figure supplement 3). Due to the predominance of rods and abundance of rhodopsin, staining of outer segments using such methods usually results in an overwhelming signal that confounds resolution of individual cells and their features of interest (here outer segments). Since we had to label the electroporated cells anyway, we directly quantified the IS+OS lengths of mGFP+ photoreceptors. However, the reviewer’s suggestion is useful in highlighting the cell-autonomous nature of our perturbations. We do not observe a decrease in the length of rhodopsin-positive outer segments of photoreceptors that did not receive the plasmid (mGFP- photoreceptors in the electroporated patch) when compared to outer segments outside the electroporated patch. We thus concluded that reduction in OS length of mGFP+ photoreceptors was due to a cell autonomous effect. One would expect that an effect on Mueller glia or the RPE would have affected the non-electroporated rods in the patch as well. In addition, this also addresses the reviewer’s concern that a cell receiving the knockdown construct does not result in an adverse phenotype on its nonelectroporated neighbor. Besides this suggested experiment, we are also presenting phenotypes observed using sparse electroporation of the retina, which is achieved by diluting the plasmid. Fewer cells are electroporated in this case, and again we observe a reduction in outer segments of electroporated photoreceptors.

Changes in the manuscript: Added Figure 1—figure supplement 3. The text is amended to describe this figure (subsection “Active LDHA supports outer segment biogenesis”, first paragraph). Added Figure 3—figure supplement 3 with amends to the text body (subsection “Nonequivalent roles of pyruvate kinase isoforms”, second paragraph) for sparse electroporation. Description of rhodopsin antibody is included in Supplementary file 3.

9) For the experiments in Figure 1J,K and also other experiments where outer segment and inner segment length were measured – were all the retinas collected at the same time of day?

We appreciate pointing out this aspect to us. We did consider a possibility of a circadian/diurnal effect on the OS length. The retinae from six sets of animals: control (light and dark), LDHAsh (light and dark), PKM2sh (light and dark) were harvested in the first half of the day and within 3 hours of each other. Similarly, efforts were made to collect the AAV-TIGAR and AAV-mGFP infected retinae for lactate analysis at the same time of the day, as they became available. The timing was inadvertently not incorporated in the earlier draft. We have now updated the Materials and methods sections (” Mice and animal husbandry” and “ATP, Lactate and NADPH assay”).

10) Maybe I overlooked it, but I could not find the part of the text that refers to Figure 2—figure supplement 1 panels C-F.

The text referring these specific figure panels is in the second, third and fourth paragraphs of the subsection “Allosteric regulation of glycolysis in photoreceptors”.

11) Subsection “Nonequivalent roles of pyruvate kinase isoforms”, first paragraph: Please also cite recent paper from the Rajala lab (PMC5121888) that show expression patterns of PKM1 and PKM2 and tyrosine phosphorylation of PKM2. Also please cite the relevant IOVS paper by Casson et al. PMID: 26780311. Another paper that should be cited regarding PKM2 in the retina is Molecular Vision paper by Rueda et al. (PMID: 27499608)

These references are now cited.

12) The difference in the overall thickness of the retina is particularly obvious in Figure 3—figure supplement 1G (the rescue with huPKM2). It seems like the whole retina is affected by the PKM2 expression. Is that correct? The authors should address the possibility that the whole retina is affected in a non-cell autonomous way. (Minor note: the legend refers to the boom panel on the right but it is on the left).

The reviewer is correct in pointing out that all the retinal layers are affected, including the INL. Though that panel was more towards the periphery (which accentuated the retinal thinning), we have replaced it with a panel that lies in the central retina, closer to that of the rescue panels. We have observed this phenotype (thinning of ONL and INL) reproducibly in many different retinae that get well-electroporated. PKM1+2sh knocks down both PKM1 and PKM2 and is more effective in knocking down PKM2 than the PKM2sh (see the western blots in Figure 3—figure supplement 1, Panels D and E in the revised draft). Thus it is likely that this construct could perturb retinogenesis to a greater extent and/or can also affect the cells in the INL (that express PKM1). The genetic sufficiency of PKM2 in restoring the OS length is key and forms an important aspect in concluding the nonequivalent roles of PKM1 and PKM2 in the retina. We appreciate the reviewer’s comments on the labeling of panels. We have now corrected this in the figure legend.

13) Subsection “Nonequivalent roles of pyruvate kinase isoforms”, fourth paragraph: PKM2fl/fl and Rod-cre mice were used to knockout PKM2 in rods for the lactate measurements. Please note whether the rods were shorter in these experiments. Also, if I'm right that the overall thickness of the retina changes do that also occur in the PKM2fl/fl/rod-cre retinas?

We are including the panel for rhodopsin staining to label the rod outer segments of Rod-cre> PKM2fl/fl in Figure 3—figure supplement 2H (Figure 3—figure supplement 1H in earlier version). Fortunately, the mosaic nature of recombination due to the Rod-cre line enables us to compare RHO staining in the region where recombination of the M2 exon has not occurred (lack of PKM1 staining) with the regions where the recombination has occurred (appearance of PKM1 staining in the rods). At this time point, one can appreciate a small decrease in RHO+ outer segments in the recombined region and a slightly longer outer segments of the rods that do not show PKM1 expression in the same retina. We also wanted to assess if PKM2 deletion makes the photoreceptors vulnerable and reduces their survival over the period of time. The data from 8-month-old animals is included in the revised draft. We also examined aged mice (74-85 weeks old) for surviving photoreceptors by plotting number of rows of photoreceptor nuclei from the center of the retina (optic nerve head) to periphery for the typical spider plots. We did not observe an apparent difference between PKM2fl/fl; Rod-cre retinae and Bl6/J age-matched controls. Due to the mosaicism of the Rod-cre line used in our experiments, we may not have been able to detect a phenotype that perhaps would be more obvious in the full ko. Increasing the number of aged animals or using alternative photoreceptor Cre lines are approaches that we are currently undertaking. Using Cre lines that exhibit less mosaicism (than the currently used Rod-cre mouse line) might have an added advantage of giving more penetrant metabolic phenotypes with LDHA or PKM2 conditional animals. Since these results would not change our conclusions about the role of LDHA and PKM2 in the glycolytic preference of photoreceptors, we are not including the results of those experiments in the interest of the amount of time it would take to breed and then age such animals.

14) In the subsection “Fibroblast growth factor signaling regulates anabolism”, the authors should cite a recent paper, PMC5121888, that also confirms that Y105 of PKM2 is phosphorylated in mouse retinas.

The work by Rajala et al. (2016) is now cited. Their conclusions derive from detection of the signal using IHC and western blot using the phospho-Y105-specific antibody. They could detect the signal in photoreceptors on IHC (that predominantly express PKM2) as well as in the IPL, as well as in what appears to be Mueller glia (all of which would arise from PKM1 and not PKM2). They also see an increase in signal in all of these locations in a light and activity-dependent manner. Of critical importance to such an approach is the establishment of the phospho-specificity of the antibody, and ruling out an apparent increase in signal intensity due to an increase in the total protein itself. To their credit, they have been able to observe this phosphorylation on IHC and we hope that the issue of phospho-specificity was addressed during peer-review, though we did not come across this aspect in the paper itself. If the phospho-specificity of the anti-Y105 antibody on IHC is indeed established according to the authors’ own data it would indicate PKM1 is regulated in a similar manner as PKM2. On the other hand, the caveat with using whole retinal lysates is that a likely scenario of PKM1 phosphorylation at Y105 cannot be ruled out as a confounding factor (as we highlight in Figure 4A and which is possible as shown by phosphorylation of PKM1 in muscle lysates in 4B).

However, our results align in many respects and we highlight both of these aspects in the text (subsection “Fibroblast growth factor signaling regulates anabolism”, first paragraph).

15) Please include MW marker positions on Figure 4B. Also, please explain what the "isotype-matched" means.

The marker positions are now included. The concern of the reviewer regards the apparent molecular weight similarity of PKM2 (~56 kDa) with IgG heavy chain. Our method of immunoblotting (following immunoprecipitation) using a conformation-specific secondary antibody eliminates a likelihood of obfuscating bands due to denatured immunoglobulin chains.

An isotype-matched rabbit monoclonal (IgG) was used as a control for non-specific binding in IP experiments using PKM2 rabbit monoclonal. This is now explained in the figure legend.

16) The authors should be careful to not over-interpret data in some of their descriptions. For example, the authors state "Impact of glycolytic perturbation on nucleotide availability was directly visualized…". Since FGF signaling can influence other processes besides glycolysis I think this is over-interpreting the data. It would be better to just say that FGF receptor signaling influences nucleotide availability and this could be linked to limitation of glycolysis. Alternatively the connection could be strengthened by evaluating the effects of PKM2 or LDHA inhibition on EU incorporation.

The advantage of the EU incorporation assay was that it enabled us to directly visualize nucleotide incorporation in nascent RNA in photoreceptors. In our opinion this assay is much better than other metabolite assays where whole retinal lysates are needed. We stand by our claim but have modified the sentence to incorporate the reviewer’s suggestion. We do show the effect of Ldha inhibition by oxamate. We agree with the reviewer that FGF inhibition has a much stronger effect than oxamate itself, which is likely due to the role that protein tyrosine phosphorylation plays at multiple nodes (LDHA, PKM2 and Pyruvate dehydrogenase kinase) in regulating the glycolytic metabolism in addition to other nonglycolytic roles that FGF signaling might have. We have tried multiple effectors of PKM2-specific activity (DASA, TEPP and ‘Compound 9’). All of these have been shown to significantly affect aerobic glycolysis in cancer. But in our hands these agents did not change the lactate production from the retina, the quaternary state of retinal PKM2, or nucleotide incorporation in photoreceptors. We are not certain if the differences are due to the efficacy of these drugs (though we used three different drugs at a range of concentrations) or some innate differences in PKM2 protein milieu or its regulation in photoreceptors and cancer cells that may underlie differences in retinal and cancer aerobic glycolysis.

Reviewer #2:

Vertebrate photoreceptors are among the most metabolically active cells, exhibiting a high rate of ATP consumption. This is coupled with a high anabolic demand, necessitated by the diurnal turnover of a specialized membrane-rich organelle, the outer segment, which is the primary site of phototransduction. It is not clear to date that how photoreceptors balance their catabolic and anabolic demands. The current study has attempted to address this fundamental issue in photoreceptor biology. The authors have used several biochemical, immunological, genetic and viral transduction methods to address the importance of glycolysis on the outer segment biogenesis. The study is interesting but failed to support the authors claim that glycolysis regulates the outer segment biogenesis.

The major criticism of the reviewer is “The study is interesting but failed to support the authors claim that glycolysis regulates the outer segment biogenesis.” Our conclusion from the data differs from the referee’s summary of our work for primarily two reasons:

1) Our approach of uncovering reliance on glycolysis utilized functional perturbations and assessments of three different nodes of glycolysis, establishing genetic necessities for key players, metabolic nonequivalencies of glycolytic enzyme pairs (Ldha vs Ldhb and PKM1 vs PKM2) and genetic dissection of allosteric control of photoreceptor glycolysis by metabolic complementation. We have consistently observed dependence of rod outer segment length on a functional glycolytic pathway.

2) “[…]glycolysis regulates the outer segment biogenesis.”: We have refrained from a general statement that would implicate a metabolic pathway of broad survival significance as a controlling mechanism in biogenesis of a specialized organelle. On the contrary, we have posited that the multiple metabolic demands posed by outer segment shedding and a need to replenish it, require a clever approach on part of cells. The summary of our findings is that photoreceptors rely on aerobic glycolysis which they themselves carry out and that this pathway is subject to regulatory strategies at multiple levels. We show that perturbing these regulatory steps results in visibly shorter outer segments and reduced steady state concentration of metabolites that are necessary for anabolism. We have not concluded that glycolytic pathway can regulate the complex aspects of OS biogenesis, but instead infer that photoreceptors rely on a functional glycolytic pathway to enable OS growth.

1) In these studies, authors have examined the isoform expression of LDH, PKM2, then phosphorylation state of PKM2 under dark- and light-adapted conditions, identified FGF signaling promotes PKM2 phosphorylation and splicing regulation of PKM1 and PKM2. Some of these studies have been done before by other labs (see below), which the authors did not acknowledge or reference in their manuscript. These include:

• Identification of LDH isoforms (Casson RJ et al. 2016 IOVS).

• PKM1 and PKM2 isoform characterization, light-dependent tyrosine phosphorylation of PKM2 (Rajala et al. 2016 Sci Rep).

• The authors claimed in this study that fibroblast growth factor (FGF) signaling was found to regulate glycolysis through phosphorylation of PKM2. This finding is not novel as it has been elegantly shown in tumor cells that FGFR regulates PKM2 phosphorylation (Hitosugi et al. 2009 Sci Signal).

• Regulation of PKM1 and PKM2 splicing (Su et al. (2017 Mol Cell Biol)

The authors are urged to cite these earlier references and give proper credit for these studies. They could discuss how the published results are similar or differ from their observations in this manuscript.

We are addressing the 4 specific issues raised by the reviewer here in comment#1 itself, and not later if they appear redundantly in comments 2 through 16.

“Identification of LDH isoforms (Casson RJ et al. 2016 IOVS).”: Casson et al. have not identified LDH isoforms in the retina. They have reported Ldha expression in the retina by western blot and IHC in multiple species. However, missing from their analysis and critical for the interpretation of significance of their work in terms of glycolysis, is the expression pattern of Ldhb. This is especially important because before them Michael Kalloniatis’ group has reported that, in the retina, Ldha and Ldhb subunits can assemble in all 5 possible combinations (Acosta et al., 2005. PMID: 16163270) underscoring the importance of careful analyses of all isoforms in cells of interest. Secondly, we do not endorse the assumption that mere expression of enzymes implicated in cancer signify a similar anabolic dependency. The PK isoforms studied by Casson et al. and Ldha are known to be expressed in many other cell types and not just in cancer cells. For example, skeletal muscle expresses Ldha, but does not carry out aerobic glycolysis. On the other hand, Ldhb can promote a Warburg-like phenotype as well (Oginuma et al., 2017. PMID: 28245921). Their work provides a descriptive catalog of expression of PKM1, PKM2 and Ldha in many mammals and we acknowledge that. But we disagree with the reviewer’s specific assessment of Casson et al.’s work.

“PKM1 and PKM2 isoform characterization, light-dependent tyrosine phosphorylation of PKM2 (Rajala et al. 2016 Sci Rep).”: We had not come across Rajala et al.’s and Rueda et al.’s work when we were preparing our manuscript. It was inadvertent omission on our part. We are now citing these papers. Though already addressed as an issue in response to reviewer#1, we would like to reiterate that although we are glad to find similarities in the manner in which tyrosine kinase signaling is regulating PKM2, Rajala et al.’s results should be considered as PKM regulation as opposed to PKM2. We are discussing it in more detail in the text.

“The authors claimed in this study that fibroblast growth factor (FGF) signaling was found to regulate glycolysis through phosphorylation of PKM2. This finding is not novel as it has been elegantly shown in tumor cells that FGFR regulates PKM2 phosphorylation (Hitosugi et al. 2009 Sci Signal).”: Our interpretation of work by Hitosugi et al. differs from that of reviewer’s. Hitosugi et al. demonstrated direct phosphorylation of PKM2 by oncogenic forms of FGF receptor 1. They model a translocation event where the C-terminal kinase domain of the FGF receptor is fused with a self-association motif of ZNF198, rendering the kinase constitutively active. The cells stably expressing this fusion protein lose their dependence on interleukin 3, which is observed in the parent cell line. Thus, it is not FGF signaling per se but direct phosphorylation by the kinase domain of FGFR1 that is responsible for PKM2 phosphorylation. The authors show direct binding and phosphorylation of PKM2 by FGFR1. These authors have also shown constitutively active versions of other tyrosine kinases can also result in PKM2 phosphorylation at Y105. This varies significantly from the reviewer’s interpretation. Secondly, our intention was not to discover a novel aspect of PKM2 regulation by tyrosine kinase signaling. We wanted to know the specific pathway that regulates PKM2 tyrosine phosphorylation in the retina. To our knowledge, this signaling pathway was not known to interact with PKM2 in the retina. Thirdly, it is not only PKM2 that is regulated by FGF signaling. As Figure 4E shows, Ldha is also a target of this pathway. Subsequent panels in Figure 4 aim to demonstrate the role of FGF signaling in regulating aerobic glycolysis in the retina, which we believe are exciting data and worth exploring further.

“Regulation of PKM1 and PKM2 splicing (Su et al. (2017 Mol Cell Biol)” and comment (13) from below “Figure 3—figure supplement 2 does not add any new information. The authors' data show opposite expression of these splicing factors. There was a study recently published showing that RBM4 Regulates Neuronal Differentiation of Mesenchymal Stem Cells by Modulating Alternative Splicing of Pyruvate Kinase M (Mol Cell Biol 2017).”: We believe that these comments represent drastically different interpretation of our findings and those of the existing literature. To our knowledge the molecular basis of biased expression of PKM1 and PKM2 isoforms in the two different layers of the fully differentiated retina is not known. To us the simplest explanation was that it could be attributed to splicing factor expression, which in turn would dictate the bias and explain how postmitotic photoreceptors keep expressing PKM2. While expression of SRSF3 aligns with the existing model for preferential inclusion of M2-specific exon, we expected PTBP1 expression to follow a similar pattern, but it was not. Thus the reviewer’s implication that somehow the mutually “opposite” expression of these splicing factors is expected goes against how we interpret the literature. Secondly the reference provided by the reviewer (Su et al., 2017) (that was published a week before our submission) looks at RBM4 splicing factor-dependent PKM1 expression by antagonizing Ptbp1. This reference still does not explain our observation that Ptbp1 is enriched in the cells that preferentially express PKM1 in the retina. The RBM4 paper would predict that Ptbp1 splicing activity in the INL would be attenuated, in order to promote PKM1 expression. We still remain puzzled by the observation that PTBP1 is expressed at higher levels in the INL. A relevant reference for retinal splicing that touches on PTBP1 is the work by Murphy et al., (PMID: 2754135), although their work still does not explain how PKM2 is expressed in photoreceptors. Again, the splicing choice of photoreceptors is peripheral to our primary focus on glycolysis, but we thought of highlighting the discrepancy between existing splicing models and our observation by providing these data on the expression of the splicing factors.

2) LDH isoforms identification has previously been reported (Casson RJ et al. 2016 IOVS). The authors must cite this manuscript.

Addressed above.

3) The authors stated in the manuscript (subsection “Lactate producing isoform of Ldh in photoreceptors”, last paragraph) the recombination efficiency with rod-cre varied between 50-90%, but that is not correct. The rod-cre used in these studies will not recombine more than 50%. The authors have shown only protein expression by Western blots. They need to show the deletion by immunohistochemistry.

We show that the cre line is mosaic and in the sample section shown in Figure-1, figure supplement 2, there are more than 50% rods that show recombination history. Is the reviewer stating that the recombination was maxed out at 50% from his/her experience? It is worth noting that expression of engineered alleles in mice can be variable even in inbred strains.

4) The authors have shown that lactate production was significantly reduced in conditional LDH-A mice. Why did the authors not study OS biogenesis in LDH-A deleted mice? The shRNA strategy is not well justified. Have the authors examined the OS in LDH-A KO mice? Generally, the shRNA approach may not knock-down completely the gene of interest but conditional deletion will? The authors observed the shortening of OS. Could this be an off-target effect? LDH-A is also expressed in other layers of the retina (INL and IPL).

Immunohistochemistry is not the ideal way to demonstrate OS length; hence the authors measured IS+OS. The authors should use ultrastructural studies, such as EM or high resolution LM to demonstrate the OS length phenotype. Some of the micrographs show thinning of the outer nuclear layer thickness (e.g., DAPI stained sections in Figure 1H, 1J, 2C, and 3C, suggesting retinal degeneration in these genetically modified retinas, which could argue against shortening of OS length. Did the authors do TUNEL or any other test for dying photoreceptor cells?

Though we have already stated our reason to use electroporation and not rely on knockout for cellular phenotypes, we are stating them here as a response. Our primary concern was to establish the cell autonomous requirement of aerobic glycolysis for the photoreceptors and hence we chose an electroporation-based approach. Our major concern with knockouts affecting large numbers of cells within a specific area was that it might change the metabolic environment that the photoreceptors sit in. This would cloud our interpretation of the autonomy of a photoreceptor phenotype, as it might reflect indirect effects on Mueller glia and/or RPE, etc. Another possible problem with a large swath of cells knocked out is that, if we did not observe a phenotype, it could be due to a cellular adaptation to an altered environment. This concern has recently been highlighted by the work of Martin Friedlander (Kurihara et al., 2012. PMID: 23093773) who show that in the face of catastrophic metabolic alterations brought about by ablation of the choriocapillaris, the rods are able to persist for months and maintain a functional response, indicating remarkable adaptability. Also a gene knockout might not necessarily recapitulate the phenotype observed with acute interference such as RNAi (e.g. Doublecortin and TUG1 give phenotypes with sh knock down but not in full KOs). Though the concerns usually stem from functional compensation in case of whole body knockouts, one cannot effectively rule out a similar effect in conditional knockouts as well. Additionally, electroporation enables us to address issues that cannot be addressed via classical conditional alleles, such as establishing that catalytic activity of Ldha is essential (rather than the entire protein itself), allowing us to devise an experimental strategy to target an allosteric metabolite (F-2,6-BP) which would not be otherwise straightforward due to three genes (PFKFB1, PFKFB2 and PFKFB4) (issue of functional redundancy and ruling out pleiotropic effects) expressed in photoreceptors, non-equivalent roles of PKM1 and 2, and Ldha and b.

Regarding off-target effects of shRNA, we showed that we could complement the shRNA phenotype by a non-targetable construct of the targeted gene. This is the gold standard control for specificity.

We question the usefulness of EM ultrastructural studies for the primary focus of our study. First, it would be difficult to pinpoint the affected photoreceptor outer segment amidst other normal looking OSs. Secondly, our method of coelectroporation with mGFP allows one to reliably assay the shortening of outer segments, which we believe is sufficient for assessing the effect of glycolysis. Though there are advantages offered by EM studies, these are tangential to our goals of establishing the function of a metabolic pathway for photoreceptor cells. However, such studies would be well suited for understanding the cellular responses at organellar/suborganellar level, which we would very much like to know.

Thinning of the ONL is also observed in dark-reared sh transfected retinae, but there is a rescue in terms of OS length. Thus retinal degeneration cannot be a causative factor for OS shortening. In addition, the non-electroporated neighboring rods have outer segments that are identical to those from rods outside the electroporated patch. We have addressed this issue in the Discussion section. These data indicate that degeneration is not responsible.

We have done TUNEL staining of these retinae. We did not observe TUNEL positive nuclei at the timepoints shown in the manuscript. However, we cannot rule out the occurrence of cell death since we do not know how long a dead cell persists in the adult retina in a state capable of being captured by TUNEL. Given that the cells would die asynchronously, and there was not a large number of cells missing, we estimate our chances of detecting a TUNEL-positive nucleus in a given retinal section to be very low, even if there was death.

5) Could shRNA knock down in other retinal layers (may be Muller cells) may indirectly affect the structure of OS?

We have included rhodopsin staining of nonelectroporated cells. In the case of non-autonomous effects, these cells would have been expected to have been affected, but this was not the case. We discuss this issue in the Discussion. Also to note is the experiment with sparsely electroporated PKM2sh retina, which further reduces the possibility of non-autonomous effects.

6) Figure 1K – There is no evidence of disc shedding in this experiment. Could there be less opsin trafficking to the OS? Such a possibility cannot be ruled out. This experiment is overstated.

We do not claim it to be disc shedding (“[…]presumably due to less disc shedding). We have incorporated the suggestion of reference for this effect made by another reviewer. Our understanding based on the literature is that opsin trafficking is necessary for OS biogenesis, and thus we did not consider less opsin trafficking to be a realistic event that could rescue the photoreceptors in dark.

7) The statement "lactate production by the photoreceptors cannot be attributed to lack of mitochondrial activity." The authors have done experiments on (Figure 1L) wild- type retinas. Have they carried out these experiments in LDH-A-knockdown or KO mice?

We have now included cytochrome oxidase assay on LDHA KO mice (Figure 1—figure supplement 4). We do not observe a decrease in COX activity after LDHA deletion.

We would like to reiterate that the purpose of assaying for mitochondrial activity in wild type retina was to examine if a cause-and-effect relationship exists between LDHA expression and mitochondrial function. We observed that despite having normal functioning mitochondria, photoreceptors have LDHA expression. Thus, LDHA-dependent aerobic glycolysis seems to be a choice as opposed to a consequence of poor mitochondrial activity. Thus, photoreceptor lactate production fits the definition of aerobic glycolysis.

8) Figure 2A and 2F – what is the rationale to regulate TIGAR expression spatially and temporarily? In Figure 2F the authors used AAV-mediated expression of TIGAR. Lactate levels were done in AAV-TIGER (Figure 2G) but not for Figure 2A? It is very confusing, and there was no rationale provided for these experiments. It seems that authors may have difficulty in measuring lactate levels for the inducible expression system?

The rationale is already stated in the text (subsection “Allosteric regulation of glycolysis in photoreceptors”, last paragraph). We understand the reviewer’s concerns that the multiple reasons for using TIGAR overexpression could be confusing. We have modified our description of this strategy in an effort to improve the clarity of our rationale.

9) Figure 3A – PKM1 and PKM2 expression has recently been reported (Rajala 2016 Sci Rep). The authors have not cited this reference.

As stated before, this was an oversight on our part. We did cite work by Lindsay et al. (2014) who first characterized these isoforms in the retina, but not every subsequent work that follows theirs’ unless the data differed from ours or the work provided insights that could not be obtained from Lindsay et al. Though, to our knowledge, the first evidence of PKM2 in the retina is attributable to the work of Santa Ono and colleagues (Morohoshi et al. 2012. PMID: 22465421).

10) The authors have shown the developmental expression of PKM2 and PKM1 on western blots (Figure 3B), which is not the ideal way to show the developmental expression. If the authors wanted to show this, they should provide immunohistochemistry or ISH.

The developmental milestones at indicated time points in the postnatal retina are already well established. Since the increase in PKM1 expression is very gradual and we did not observe a remarkable change in expression level at any of the time points, we did not think it warranted further in-depth analysis. This manuscript is focused on the adult retina and we included the time course data simply out of interest.

11) The authors used rod-cre to delete PKM2 and measured LDH activity (Figure 3G). For structural studies, they used shRNA and examined OS length (Figure 3G). These studies are not convincing. Why did they not observe similar shRNA effects with conditional PKM2 KO mice?

We would like to clarify that we did not measure LDH activity in these retinae as the reviewer suggests. The rationale for using shRNA is already described previously. We do observe a slight reduction in outer segment length in PKM2 conditional retinae very early. With progressive age the OS length and the condition of the retina worsens.

12) There is no indication of how much PKM2 is deleted or knocked down. In the absence of these experiments, it is very difficult to interpret the data. Moreover, the authors did not carry out any functional studies, such as ERG to examine the role of PKM2 in photoreceptor functions?

The knockdown of PKM2 and steady state protein levels were already depicted in Figure 3—figure supplement 1 (C, D, E) (Figure 3—figure supplement 2C,D,E in the revised manuscript).

13) Figure 3—figure supplement 2 does not add any new information. The authors' data show opposite expression of these splicing factors. There was a study recently published showing that RBM4 Regulates Neuronal Differentiation of Mesenchymal Stem Cells by Modulating Alternative Splicing of Pyruvate Kinase M (Mol Cell Biol 2017).

This comment is already addressed before.

14) The authors stated that PKM2 deletion upregulates PKM1, but has no effect on photoreceptor structure (Figure 3—figure supplement 1H). On the other hand, forceful expression of PKM1 had a reduction in the length of OS? How do authors explain this discrepancy?

In rods where recombination occurred, and expression of PKM1 resulted, slightly smaller outer segments occurred, compared to non-recombined neighbors in the same retina. The initial characterization of the PKM2 conditional allele (Israelsen et al., 2013) has already shown that deletion of the M2-specific exon results in appearance of PKM1, but the expression levels are rather low (~40% of the PKM transcript are PKM1 and PKM protein steady state suggests lower expression level than the non-recombined cells). Thus, we speculate that the PKM1 expression levels from the two genomic loci in the conditional knockout never match the higher levels achieved by PKM1 overexpression driven from multiple copies of plasmid using the strong CAG promoter.

15) Figure 4 – FGF signaling – Authors have identified that FGF signaling promotes the phosphorylation of PKM2. It is not a novel finding. It has been shown in tumor cells (Hitosugi et al., 2009). The authors have not acknowledged this information in the current manuscript.

This comment has been addressed before.

In rods where recombination occurred, and expression of PKM1 resulted, slightly smaller outer segments occurred, compared to non-recombined neighbors in the same retina. The initial characterization of the PKM2 conditional allele (Israelsen et al., 2013) has already shown that deletion of the M2-specific exon results in appearance of PKM1, but the expression levels are rather low (~40% of the PKM transcript are PKM1 and PKM protein steady state suggests lower expression level than the non-recombined cells). Thus, we speculate that the PKM1 expression levels from the two genomic loci in the conditional knockout never match the higher levels achieved by PKM1 overexpression driven from multiple copies of plasmid using the strong CAG promoter.

16) Figure 4C – PKM2 undergoes a light-dependent tyrosine phosphorylation on Tyr105. These studies have recently been reported by Rajala et al. 2016 (Sci Rep). The authors have not acknowledged this study.

We have addressed this comment above.

Reviewer #3:

This manuscript provides novel and interesting data on the reliance of aerobic glycolysis for photoreceptor outer segment renewal. Overall, the paper is very good and a significant contribution. However, there are some significant problems that need addressing before the results and conclusions that are presented can be accepted. In addition, there are several additional items that are off putting and overstepping the presentation and results.

First, the authors are only addressing aerobic glycolysis in the rod inner segments and outer segments. There does not appear to be any data for cones or rod and cone photoreceptor synapses. So, the title should reflect this and state something akin to "Aerobic glycolytic reliance promotes anabolism in rod photoreceptor outer and inner segments."

Our work does not address aerobic glycolysis in a given cellular compartment, but uses the readout of outer segment length as an assay of perturbations of aerobic glycolysis. The perturbations might affect other cellular compartments, but our assay has been only of outer segment length, except in the case of acute FGF inhibition, where we saw an effect on nascent RNA synthesis in the nucleus. But given that this work highlights the propensity to preferentially carry out glycolysis (glycolytic reliance) in order to maintain outer segments, nucleotides and NADPH (anabolism) in photoreceptors, we believe that the current title is accurate. Further studies might uncover if the metabolic effects of specific perturbations vary among specific compartments.

Second, the authors should provide more detailed and important information in the Methods. This information is critical for proper interpretation of the results. For example, what the light level in the animal room and cages, what time of day were the mice sacrificed, what area and quadrant of the retina was quantified for OS-IS measurements, how was this location established, what are the measurements for OS and IS alone as they could both change, how many non-adjacent sections per retina were quantified per mouse, what was the volume of incubation for the lactate assay, what type of confocal was used and how many Z-stacks were there per image, what was the exact age of mice used for the histology as in the subsection “Dissections and adult explant cultures”, the authors stated that they used P23-P28 mice. According to LaVail 1973 J Cell Biol, the outer segments in the C57BL/6J mice reach their rate of synthesis and disposal at around P21-25.

- Light level: This information is now provided

- Time of the day of harvest: Most of the animals were euthanized for tissue harvest between 3-9 hours after lights were turned on. Specific time points are reported for experiments of dark-reared and light-reared animals and AAV-infected retinae for lactate assay (where we wanted to avoid small differences due to possible circadian/diurnal changes). Whenever possible, animals belonging to the same assay group were sacrificed together. All data points are presented, i.e. we do not exclude any outliers.

- Quadrant of the retina: Electroporation of the retina yields a patch of transfected cells. The exact shape and size of the patch varies among animals, though roughly in the same place. The extent to which the DNA will spread in the subretinal space and how many cells will get transfected cannot be predicted. In the experiments presented here, the injections targeted the dorso-nasal quadrant of the right eye, as a matter of experimenter preference, though in some cases the transfected cells may be from a nearby area.

- OS and IS measurements: In an ideal situation we would have quantified these two compartments separately. However, we resorted to IS+OS quantification in two possible situations:

i) In cases where a clear boundary between IS and OS was lost

ii) In cases where the OS was present but too small to accurately measure by itself

- Number of nonadjacent sections: For each electroporated retina that was sectioned on multiple slides, we chose the section that had the biggest electroporated patch. We aimed to quantify all the photoreceptors in this patch, which involved multiple nonoverlapping visual fields, by moving along the x-axis in the electroporated patch. This approach samples across a large area and thus can be considered equivalent to a nonadjacent section which would sample along the z-axis rather than along the x-axis.

- Type of confocal and number of Z-sections: Confocal type is already provided in the Materials and methods subsection “Immunohistochemistry”. There was a typo in the model number (LSM710 was mentioned as LSM10). It has been corrected. This information now also includes another confocal that was used for some experiments in the revision. The number of Z-sections varies between individual sections and is governed by the number of electroporated cells and their location in the tissue slice. For some sections we could sample all the photoreceptors in 12 slices, while for others we did as high as 20 slices.

- Exact age of mice used in subsection “Dissections and adult explant cultures”: First we would like to clarify that we have used these mice for explant culture and not histology for outer segments. Though we are not certain of the reviewer’s concern that the choice of age-range would pose for metabolic assays, we speculate that he/she is concerned about changes in the metabolic capacity of the photoreceptors. As stated by the reviewer, most photoreceptors achieve the adult synthesis and shedding rates by after 21 and 25 days. According to those data, between 23 and 25 days the mean OS length increases by 10% in length with ~70% overlap in the range of data. We have quantified lactate secretion from fresh untreated retina many times (from what would essentially cover the entire age range mentioned in the manuscript) in multiple different experiments. We went back and examined if lactate secretion was lower in experiments when we used 23-25 day old mice. There does not seem to be a trend or significant difference to indicate that a difference of two/three days in that age group would make in glycolytic capacity. In addition, the siblings from the litter (or litters of the same age) were used in an experiment that also included controls. Thus we believe that age differences for any data points that come from mice 23 or 24 days old should not contribute to a statistical differences.

- Volume for lactate assays: 0.5 mL. Can be found in the Materials and methods subsection “Dissections and adult explant cultures”.

Third, the authors conclude from looking at their retinas 42-45 days following electroporation that the shortened rods were due to decreased synthesis due to decreased aerobic glycolysis. This cannot be accepted at face value without conducting two additional experiments: show rod disc synthesis results [e.g., per Young and Bok studies] and rod phagocytosis results [e.g., per LaVail studies].

We conclude that OS length is dependent on functional glycolytic pathway. By three separate genetic perturbations targeting the glycolytic pathway at three different nodes, we consistently observe that the OS length reduces. The work by Richard Young and Matthew LaVail has indeed provided experimental evidence that has laid the groundwork for studies on OS biogenesis. The work utilizes EM ultrastructural studies. While those assays give unbeatable resolution, they are either not compatible or incredibly difficult to do when merged with our approach of electroporation and isolated cell labeling. Even if we are able to successfully implement those methods with our assays, the information on the dynamics of OS biogenesis after glycolytic perturbation can be inferred but the final OS length (which is an important end result) will still need to be assessed. For the latter part, confocal microscopy is reliable and sufficient.

Fourth, please explain why eye opening occurred on day 11 in these studies. Most published mouse paper find that this event occurs at postnatal day 14 +/- 1 day.

We followed our mice every day to assess eye opening so as to ensure that the animals were not exposed to unwanted amounts of light. We were concerned that the eye of the electroporated animal might open earlier than usual because of the incision at P0 made to expose the eye for subretinal injections. We found the eyes opened ~postnatal day 11, whether they were injected or not. These mice were transferred to dark, as reported. Author response image 1 shows an injected animal at P11 and an uninjected animal at P12. As can be seen in both the images, the eyes are open. To our knowledge, the beginning of eye opening is also used as guide for staging mice at P11 (Please refer the JAX mouse staging chart at https://oacu.oir.nih.gov/sites/default/files/uploads/training-resources/jaxpupsposter.pdf)

Fifth, the choice of words in several places throughout the manuscript are overstated or not properly used. This often results from lack of citation of previously conducted work. For example, they state that "The cell types that carry out aerobic glycolysis in the normal adult retina have not been determined." This is just false. The work of Rueda et al. published in Mol Vis in 2016 that shows cell-type glycolysis, aerobic glycolysis, high energy transferring kinases and oxidative phosphorylation in over 20 different compartments and cells in the retina. A summary figure clearly shows all of this data. However, the authors have not cited this manuscript anywhere in their manuscript.

For example, they state that "surprisingly the steady state levels of ATP…did not differ from controls." Winkler in J Gen Physiol 1981 showed that retinal ATP levels are steady under most conditions: not surprisingly since Lowry and co-workers demonstrated in the 1960s that this energy measure is the last to change even after anoxia.

Rueda et al. looked at expression of enzymes participating in intermediary metabolism. Due to the exhaustive and descriptive nature of their study, understandably there is no functional assessment of their participation in aerobic glycolysis. One can speculate about the propensities of a cell type based on expression analysis and the model does just that. Their work is a comprehensive analysis of expression and has enormous descriptive value, which will greatly benefit the field. The glycolytic pathway has seen much investigation recently, thanks to the impetus from its relevance to cancer. Based on these developments, we believe it is the right time to investigate the functional significance of aerobic glycolysis as a phenomenon and differentiate it from housekeeping glycolysis. This type of functional evaluation is required, as one cannot conclude functions from expression of glycolytic genes alone. Thus we stand by our conclusion that the cell types responsible for aerobic glycolysis are not known since we are unaware of any functional assessment of this phenomenon (not to be confused with role of glycolysis in retinal physiology).

It is important to refer the context in which we have used the statement on line 90 of earlier draft. We had observed a significant decrease in lactate production upon oxamate treatment. We anticipated a decrease in ATP levels after glycolysis inhibition. However this was not the case. The ATP level did not change despite altered lactate levels. Reducing lactate levels would have slowed glycolysis. However, slowing glycolysis did not change ATP- this was surprising. Secondly, Barry Winkler’s work (that includes the reference cited by the reviewer) does indicate that ATP levels can change in acutely treated retinae, depending on the context. For example, 10 mM KCN causes a massive and rapid drop in ATP levels (Winkler et al., 1997. PMID: 9224285). Another point to note is that our oxamate treatment occurs for hours and it cannot be considered acute, using Winkler’s definition. Finally, Barry Winkler’s work (Winkler 1981, cited by the reviewer) shows that inhibition of glycolysis by iodoacetate results in an ATP decrease. Also, comparing the effects achieved by inhibiting housekeeping glycolysis (Winkler’s IAA) with aerobic glycolysis (LDH, oxamate), one can consider the effects as surprising.

For example, in the subsection “Aerobic glycolysis in the retina”, the authors incorrectly stated that adenylate kinase (AK) synthesizes ATP. The do not acknowledge that mouse retina also expresses creatine kinase (CK). The authors need to state that CK and AK serve to regenerate the ATP by reversible reactions that respond to the law of mass action. The implication of these enzyme reactions need to be explained as these enzymes may respond under conditions of high ATP hydrolysis or raising [ADP] (E.g., see the work of Wallimann, Linton et al. 2010; Rueda et al., 2016).

We are including mention of CK along with AK as phosphotransfer enzyme systems that can maintain ATP steady states. However, this does not change the conclusion of the experiment with azide that indicates that after aerobic glycolysis attenuation, the retina is more reliant on the mitochondrial contribution for ATP level maintenance.

For example, the authors state that "presumably as a result of less shedding". This results from the process of "photostasis' as published in Exp Eye Res 1n 1986 by Penn and Williams.

We appreciate pointing to us the reference. We are amending the sentence to encompass the regulatory mechanisms.

For example, the authors state that "Our work expands the cell types where aerobic glycolysis can occur to include a mature cell type, the differentiated photoreceptor cell". As noted, the authors did not cite the prior work of Rueda et al. published in Mol Vis in 2016. This work also pyruvate competition in LDH histochemistry assays and confocal to demonstrate that Ldh-A was preferentially located in photoreceptors[…]not cited in the first paragraph of the subsection “Lactate producing isoform of Ldh in photoreceptors”, as confirmatory. So the current manuscript "confirms and definitely expands" would be the correct terminology.

We agree that Rueda et al.’s findings should have been cited where parallels exist or where they do not agree with ours. As mentioned before, it was an oversight on our part while we were preparing our manuscript. We have since made changes to cite that work. Though our findings on expression of key aerobic glycolysis players agree, our interpretation of that paper is that the authors did not demonstrate aerobic glycolysis in the photoreceptors or the importance of that process. Demonstration of Warburg effect, by definition, critically requires production of lactic acid in oxygenated condition and in presence of functional mitochondria. Rueda et al. demonstrate photoreceptor-enriched expression of Ldha. However, Ldha is also expressed in skeletal muscles, which do not carry out aerobic glycolysis (lactic acid is produced when the need for ATP generation cannot be met by supplied oxygen). Also, it is of critical importance to consider the properties of Ldh isoforms in the context of IHC used by Rueda et al. An all-Ldha tetramer is equally adept at generating lactate as well as carrying the reverse reaction i.e. oxidizing lactate to pyruvate. Clearly the direction of the reaction in a cell is determined by many factors including, but not restricted to, the rate of pyruvate generation and ratio of NADH:NAD+ and cannot be predicted by IHC alone. So, does expression of Ldha in photoreceptors necessarily mean the cell carries out aerobic glycolysis? Why would Ldha not participate in lactate utilization, as per Mueller-photoreceptor lactate shuttle? Also, they cannot exclude the absence of Ldhb in the photoreceptors. They use LDHA-specific antibody in conjunction with the pan-Ldh antibody, which supposedly detects both Ldha and Ldhb. The subcellular staining pattern, especially in the photoreceptors, is markedly different when the two antibodies are compared.

Lastly, the ldh assay cited by the reviewer needs to be considered fully. All the Ldh isoforms can oxidize lactate to pyruvate and generate reducing equivalents for the colorimetric NBT-dependent reaction. As mentioned above, even the LdhA-only isoform (aka ldh5) can catalyze the reaction with equal ease in both the directions and can generate the reducing equivalents when lactate is given as a sole substrate. Inhibition by pyruvate is a great way to ascertain Ldhb function in vitro. But as pyruvate concentration is increased during the histochemical reaction, wouldn't the Ldh strive to establish equilibrium and generate less and less of reducing equivalents in that process? Our guess is that the decrease in staining intensity in the IS (Figure 8H of Rueda et al.) might reflect that process and not necessarily inhibition by pyruvate. How does the staining look when the molar ratios for lactate and pyruvate are flipped? Does an absence of staining in that case suggest inhibition of enzyme activity (as might be the case for Ldhb) or fewer reducing equivalents generated to make the blue ppt due to more pyruvate to lactate conversion (which Ldha might still do)? In addition, the enzyme activity assay on retinal section cannot be considered equivalent of lactate production by the retina. Thus, while we do agree that our Ldha expression patterns match, we cannot attribute functional assessment of aerobic glycolysis to their work.

For example, the Abstract says that the process of photoreceptor catabolism and anabolism is "poorly understood". It should be more precise as there are many studies on these process, but not in the context of rod outer segment biosynthesis.

Though there have been ongoing studies on retinal metabolism with an underlying motivation to understand photoreceptor physiology, our understanding of how central carbon metabolism is regulated in these cells significantly lags behind that of other cell types, like pancreatic β cells, hepatocytes or even dendritic and T-cells of the immune system. Understandably, working with retinal tissue, especially for metabolic assays, poses unique challenges and we fully appreciate those and the efforts of researchers in this field. In this regard, the Abstract specifically mentions, “How photoreceptors balance their catabolic and anabolic demands is poorly understood”. Given that many genetic tools needed to probe glucose metabolism have become available within last 5 years, the question has not been adequately probed and our work intended to do just that.

Sixth, no retinal references are provided for COX and SDH histochemistry. Several previous papers conducted these experiments in rodents (SDH in rat: Hansson, Exp Eye Res 1970: COX in rat: Chen et al. Acta Ophthalmol 1989; COX in mouse: Rueda et al., Mol Vis 2016). The latter paper in adult mouse was not cited in the subsection “Functional mitochondria in photoreceptors”, although quantitative layer-by-layer COX activity was measured and presented. The authors stated that SDH activity was not different between inner and outer retina, but this does not agree with prior published results.

We are citing these references. Though these assays are fairly common ways to probe mitochondrial activity, our intention was to repeat them ourselves from the standpoint of Ldha enrichment in the photoreceptors. Regarding the reviewer’s second comment, we would like to clarify that we have not stated that SDH activity did not differ between inner and outer retina. The statement on line 190-191 (of originally submitted manuscript) is “SDH activity was not lower in photoreceptors relative to INL cells.”

Seventh, the authors state that "PKLR transcripts were not detected in retina (data not shown)". However, they were found in the retina by other investigators (see Figure 1 of Rueda et al. Mol Vis 2016). No comment is made to acknowledge the difference.

As mentioned before, we had not come across Rueda et al.’s work while our manuscript was being prepared. From the revision of this manuscript point-of-view, we could not find any statement by the authors on further validation of their microarray results by other means or the changes in their hypotheses if PKLR is found in photoreceptors. This is especially important because James Hurley and colleagues (Lindsay et al. 2014) state that they did not detect PKLR protein in the retina by western blots. It would have been worthwhile to localize these isoforms of pyruvate kinase by IHC and put in perspective of glycolytic propensities. Thus, we cannot comment on their results in efforts to avoid direct contradiction. We are adding a figure supplement to Figure 3 to show our data. The figure supplement information for this figure as well as others that are affected, have been amended. Relevant changes in the text have been made with updated figure information.

Eighth, the authors should have used a higher resolution microscopy to show specifically the changes observed in the outer segments and possible exclude or determine the possible changes in inner segments and inner segment-mitochondria due to loss of aerobic glycolysis.

We appreciate the reviewer’s suggestions. While the emphasis of this manuscript was characterization of molecular propensities of the photoreceptors, we are undertaking studies to understand the cell biological implications of aerobic glycolysis in greater detail.

Ninth, please clarify in the first paragraph of the subsection “Allosteric regulation of glycolysis in photoreceptors”, the three listed criteria points.

The criteria were conceived as a guide to design an experimental strategy for dissecting glycolytic reliance from housekeeping glycolysis. Since most cells will negatively respond when the core glycolytic pathway is inhibited, we posited that interfering with this pathway was not the correct way of uncovering glycolytic reliance. The points are elaborated below:

1) Does not ablate core glycolytic enzymes in order to avoid pleiotropic effects due to their non-glycolytic roles: It is well known that many glycolytic enzymes (such as PFK1, HK, GAPDH, Enolase) have other roles besides glycolysis. The other enzymes in the pathway might also have roles outside glycolysis that have not yet been appreciated. A classical approach using loss-of-function analysis of glycolytic genes thus has obvious pitfalls, i.e. whether the observed phenotype is really due to their glycolytic role? We wanted to avoid that.

2) Targets glycolytic node such that impact on other biosynthetic pathways such as Pentose Phosphate Pathway (PPP) would be minimal: This was an important criteria for us. PPP has important roles in lipid biosynthesis, NADPH metabolism, nucleotide metabolism and all of these have a significant impact on the biosynthetic makeup of the cell. One could argue that phenotypes arising due to ablation of an enzyme, for example hexokinase, are not necessarily due to impact on glycolysis. Attenuating a cell’s ability to phosphorylate glucose and make it available to other pathways such as PPP can also prove catastrophic. This rationale also holds true for other enzymes such as triose phosphate isomerase (TPI) i.e. does a negative impact due to TPI ablation mean importance of glycolysis or generation of methylglyoxal, a toxic molecule?

3) Uncovers glycolytic reliance and differentiates it from housekeeping glycolysis: Many studies have employed inhibition of glycolytic enzyme activities such as those of GAPDH. The relative usefulness of such an approach should be questioned and care should be taken to not attribute/extrapolate the effects of such a perturbation to aerobic glycolysis. Inhibiting housekeeping glycolysis (by inhibiting glycolytic enzymes that carry reactions from Glucose-6-P to Pyruvate) can have a detrimental effect on many cell types, not only those that carry out aerobic glycolysis.

In this regard our experiments with TIGAR satisfied these criteria. Tigar affects the allosteric regulator of a glycolytic enzyme and thus does not inhibit glycolysis. Very early on in our studies, we deemed the PFK1 node of an interest to us, because slowing down flux through this step would impact commitment of carbons for glycolysis but at the same time can potentially increase PPP, as Karen Vousden’s work has demonstrated. Since fructose-2,6-bisphosphate acts as an allosteric regulator of PFK1, targeting this metabolite satisfied the above criteria.

Tenth, Ross et al. 2010a should be 2010 as there is no "b".

We appreciate the careful review. We have now corrected the reference.

https://doi.org/10.7554/eLife.25946.030

Article and author information

Author details

  1. Yashodhan Chinchore

    Departments of Genetics and Ophthalmology, Harvard Medical School, Boston, United States
    Contribution
    YC, Conceptualization, Formal analysis, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3521-9291
  2. Tedi Begaj

    Departments of Genetics and Ophthalmology, Harvard Medical School, Boston, United States
    Contribution
    TB, Methodology
    Competing interests
    The authors declare that no competing interests exist.
  3. David Wu

    Departments of Genetics and Ophthalmology, Harvard Medical School, Boston, United States
    Contribution
    DW, Methodology
    Competing interests
    The authors declare that no competing interests exist.
  4. Eugene Drokhlyansky

    Departments of Genetics and Ophthalmology, Harvard Medical School, Boston, United States
    Contribution
    ED, Validation, Methodology
    Competing interests
    The authors declare that no competing interests exist.
  5. Constance L Cepko

    1. Departments of Genetics and Ophthalmology, Harvard Medical School, Boston, United States
    2. Howard Hughes Medical Institute, Harvard Medical School, Boston, United States
    Contribution
    CLC, Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Writing—review and editing
    For correspondence
    cepko@genetics.med.harvard.edu
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9945-6387

Funding

National Institutes of Health (R01 EY023291)

  • Yashodhan Chinchore
  • Tedi Begaj
  • David Wu
  • Eugene Drokhlyansky
  • Constance L Cepko

Howard Hughes Medical Institute

  • Constance L Cepko

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Ryan Chrenek, Lucy Evans, Parimal Rana, Lillian Horin and Alexandra McColl-Garfinkel for technical help. We are grateful to Will Israelsen and Matthew Vander Heiden for help with Pkm2fl/fl mice, Ying-Hua Wang and David Scadden for Ldhafl/fl mice and Yun Z Le for Rod-cre mice. We are indebted to Barry A Winkler for generous input on retinal metabolism, especially at the earlier stages of this work. We thank Ben Stranges (George Church lab) and Quentin Gilly (Norbert Perrimon lab) for access to the microplate readers. This work has been supported by the National Institutes of Health grant R01 EY023291 and the Howard Hughes Medical Institute.

Ethics

Animal experimentation: Animal care and use adhered to the Harvard Medical School's IACUC guidelines. Animals were handled in accordance with the protocol# 0428 and 04537.

Reviewing Editor

  1. Jeremy Nathans, Johns Hopkins University School of Medicine, United States

Publication history

  1. Received: February 17, 2017
  2. Accepted: June 1, 2017
  3. Accepted Manuscript published: June 9, 2017 (version 1)
  4. Version of Record published: July 6, 2017 (version 2)

Copyright

© 2017, Chinchore et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,010
    Page views
  • 480
    Downloads
  • 29
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

    1. Cell Biology
    2. Developmental Biology
    Nicanor González-Morales et al.
    Research Article Updated
    1. Cell Biology
    2. Genetics and Genomics
    Maja Kneissig et al.
    Research Article Updated