Aberrant upregulation of the glycolytic enzyme PFKFB3 in CLN7 neuronal ceroid lipofuscinosis

CLN7 neuronal ceroid lipofuscinosis is an inherited lysosomal storage neurodegenerative disease highly prevalent in children. CLN7/MFSD8 gene encodes a lysosomal membrane glycoprotein, but the biochemical processes affected by CLN7-loss of function are unexplored thus preventing development of potential treatments. Here, we found, in the Cln7∆ex2 mouse model of CLN7 disease, that failure in autophagy causes accumulation of structurally and bioenergetically impaired neuronal mitochondria. In vivo genetic approach reveals elevated mitochondrial reactive oxygen species (mROS) in Cln7∆ex2 neurons that mediates glycolytic enzyme PFKFB3 activation and contributes to CLN7 pathogenesis. Mechanistically, mROS sustains a signaling cascade leading to protein stabilization of PFKFB3, normally unstable in healthy neurons. Administration of the highly selective PFKFB3 inhibitor AZ67 in Cln7∆ex2 mouse brain in vivo and in CLN7 patients-derived cells rectifies key disease hallmarks. Thus, aberrant upregulation of the glycolytic enzyme PFKFB3 in neurons may contribute to CLN7 pathogenesis and targeting PFKFB3 could alleviate this and other lysosomal storage diseases. CLN7 neuronal ceroid lipofuscinosis is an inherited lysosomal storage disease typically with childhood onset of neurodegenerative symptoms. Here the authors report that in a mouse model of CLN7 disease neuronal reactive oxygen species and the activity of glycolytic enzyme PFKFB3 are increased, while PFKFB3 inhibition ameliorates hallmarks of pathology.

T he neuronal ceroid lipofuscinoses (NCLs) are a family of monogenic life-limiting pediatric neurodegenerative disorders collectively known as Batten disease 1 . Although genetically heterogeneous 2 , NCLs share several clinical symptoms and pathological hallmarks such as lysosomal accumulation of lipofuscin and astrogliosis 2,3 . Ceroid lipofuscinosis, neuronal 7 (CLN7) disease belongs to a group of NCLs that present in late infancy [4][5][6] and, whereas CLN7/major facilitator superfamily domain containing 8 (MFSD8) gene is known to encode a lysosomal membrane glycoprotein 4,[7][8][9] , the biochemical processes affected by CLN7-loss of function are unexplored, which has hampered the development of therapeutic interventions 1,10 . Forty-six disease-causing mutations are recorded in the NCL mutation database (ucl.ac.uk/ncl-disease) in CLN7/MFSD8, causing a broad phenotypic range, from classic late infantile CLN7 disease to non-syndromic retinal disease with onset in childhood or as late as the 7th decade 4 . Given that treatment for CLN7 disease is likely to be more challenging than for NCLs encoding lysosomal enzymes such as ceroid lipofuscinosis, neuronal 2 (CLN2)/tripeptidyl peptidase 1 (TPP1) 11 here we aimed to understand the biochemical processes affected in CLN7 disease. Here, using the Cln7 Δex2 mouse model of CLN7 disease, we found an aberrant upregulation of pro-glycolytic enzyme PFKFB3 in neurons that may contribute to CLN7 pathogenesis.

Results
Failure in autophagy causes accumulation of structural and functionally impaired mitochondria in Cln7 Δex2 mouse. In Cln7-null neurons in primary culture from Cln7 Δex2 mice 12 (Supplementary Fig. 1a), the mitochondrial indicators ATP synthasesubunit c (SCMAS) and heat-shock protein-60 (HSP60) colocalized with the lysosome-associated membrane protein 1 (LAMP1) (Fig. 1a and Supplementary Fig. 1b), suggesting lysosomal accumulation of mitochondria. Inhibition of lysosomal proteolysis increased the protein levels of the autophagosome marker LC3-II in wild-type (WT), but not in Cln7 Δex2 neurons ( Fig. 1b and Supplementary Fig. 1c), indicating an impairment in the macroautophagy (hereafter, autophagy) previously observed in lysosomalstorage disorders 13,14 . To assess whether this failure in autophagy affected mitochondrial turnover, SCMAS and HSP60 abundances were determined in neurons incubated with the lysosomal inhibitors. As shown in Fig. 1c and Supplementary Fig. 1d, lysosomal inhibition triggered the accumulation of SCMAS and HSP60 in WT neurons, indicating mitophagy flux 15 . However, these mitochondrial markers were already increased in untreated Cln7 Δex2 neurons and were little affected by inhibiting lysosomal function ( Fig. 1c and Supplementary Fig. 1d). In addition, PTEN-induced kinase-1 (PINK1) 63/53 ratio 16 and Parkin 17 increased in Cln7 Δex2 neuronal mitochondria ( Supplementary Fig. 1e). These data suggest that the mitochondrial clearance in Cln7 Δex2 neurons is impaired. The metabolic profile analysis revealed a decrease in the basal oxygen consumption rate (OCR), ATP-linked and maximal OCR, and proton leak in Cln7 Δex2 neurons (Fig. 1d), indicating bioenergetically impaired mitochondria. The specific activities of the mitochondrial respiratory chain (MRC) complexes ( Supplementary  Fig. 1f) were unchanged in the Cln7 Δex2 neurons. However, isolation of mitochondria followed by blue-native gel electrophoresis (BNGE), complex I (CI) in-gel activity assay (IGA), and western blotting, revealed CI disassembly from mitochondrial supercomplexes (SCs) in Cln7 Δex2 neurons (Fig. 1e). These data confirm the decreased mitochondrial energy efficiency 18 and suggest the increased formation of mitochondrial reactive oxygen species (mROS) 19 in Cln7 Δex2 neurons. Flow cytometric analysis of mROS ( Fig. 1f; see also Supplementary Fig. 1g for unchanged mitochondrial membrane potential) and fluorescence analysis of H 2 O 2 ( Supplementary Fig. 1h), confirmed mROS enhancement in Cln7 Δex2 neurons. Given the cross-talk between ROS and endoplasmic reticulum (ER) stress in disease 20 , we investigated whether Cln7 Δex2 neurons suffered from ER stress. Real-time-quantitative polymerase chain reaction (RT-qPCR) analysis of the unfolded protein response (UPR), which accumulate during ER stress 21 , showed no changes in Cln7 Δex2 neurons ( Supplementary Fig. 1i). Given that cultured neurons do not necessarily behave exactly as they do in vivo, we validated our observations in the Cln7 Δex2 mouse model in vivo. Thus, to characterize mitochondria from Cln7 Δex2 mice in vivo, we next performed electron microscopy analyses of the brain cortex ( Supplementary Fig. 1j) before and after the onset of the immunohistochemical and behavioral symptoms of the disease 12 . We found larger and longer brain mitochondria in the pre-symptomatic Cln7 Δex2 mice, an effect that proceeded with age ( Fig. 1g and Supplementary Fig. 1k), suggesting progressive mitochondrial swelling. CI disassembly from SCs in brain mitochondria (Fig. 1h) and increased mROS in freshly purified neurons from the adult brain ( Fig. 1i and Supplementary Fig. 1l) were confirmed in Cln7 Δex2 mice. Altogether, these findings suggest that Cln7 loss causes impaired autophagic clearance of brain mitochondria leading to the aberrant accumulation of structurally disorganized, bioenergetically impaired, and high ROS-generating organelle.
Increased generation of mitochondrial ROS by neurons accounts for impaired mitochondrial accumulation and hallmarks of CLN7 disease in Cln7 Δex2 mouse in vivo. Next, we assessed the impact of excess neuronal mROS on CLN7 disease progression. Cln7 Δex2 mice were thus crossed with mice expressing a mitochondrial-tagged isoform of the H 2 O 2 -detoxifying enzyme catalase (mCAT) governed by the neuron-specific 22 calcium/calmodulin-dependent protein kinase II alpha (CaMKIIa) promoter (CaMKIIa Cre -mCAT LoxP ). mCAT efficacy in vivo was previously validated 23 . The resulting progeny (Cln7 Δex2 -CAMKIIa Cre -mCAT) was analyzed and compared with littermate Cln7 Δex2 -mCAT LoxP and control (mCAT LoxP and CAMKIIa Cre -mCAT) mice. The increased mROS observed in Cln7 Δex2 -mCAT LoxP neurons was abolished in Cln7 Δex2 -CAMKIIa Cre -mCAT neurons ( Fig. 2a and Supplementary Fig. 2a), verifying the efficacy of this approach. Brain mitochondrial swelling was confirmed in Cln7 Δex2 -mCA-T LoxP mice (Fig. 2b), which also showed mitochondrial cristae profile widening (Fig. 2b), a phenomenon previously observed in cells with bioenergetically-inefficient mitochondria 24,25 . Both the mitochondrial swelling and cristae profile widening observed in the Cln7 Δex2 -mCAT LoxP mice were rescued by expressing mCAT in neurons of the Cln7 Δex2 -CAMKIIa Cre -mCAT mice (Fig. 2b). Thus, neuronal mROS participates in the accumulation of functional and ultrastructural impaired mitochondria in Cln7 Δex2 mouse brain. In line with this, the increase in SCMAS abundance observed in the Cln7 Δex2 -mCAT LoxP mouse brain ( Fig. 2c and Supplementary Fig. 2b) was abolished, or partially restored, in Cln7 Δex2 -CAMKIIa Cre -mCAT mice ( Fig. 2c and Supplementary  Fig. 2b). Brain SCMAS accumulation in autofluorescent ceroid lipopigments (lipofuscin)-containing lysosomes is a hallmark of Batten disease 2 with some exemptions 26 . Consistently with this notion, lipofuscin was accumulated in the brain of Cln7 Δex2 -mCAT LoxP mice, an effect that was ameliorated in Cln7 Δex2 -CAMKIIa Cre -mCAT mice ( Fig. 2c and Supplementary Fig. 2c). Moreover, activation of astrocytes and microglia is another hallmark of Batten disease 3 that is mimicked in the brain of Cln7 Δex2 mice 12 . We found increased glial-fibrillary acidic protein (GFAP) and ionized calcium-binding adaptor molecule-1 (IBA-1) proteins in the brain of Cln7 Δex2 -mCAT LoxP mice, suggesting astrocytosis and microgliosis, respectively; these effects were attenuated in Cln7 Δex2 -CAMKIIa Cre -mCAT mice ( Fig. 2c and Supplementary  Fig. 2d). Altogether, these findings indicate that the generation of ROS by bioenergetically impaired mitochondria in Cln7 Δex2 neurons contributes to the histopathological symptoms of CLN7 disease.
Upregulation of PFKFB3 protein and activity via a Ca 2+ /calpain/Cdk5 pathway sustains a high glycolytic flux in primary neurons obtained from Cln7 Δex2 mice. Mitochondrial ROS stimulate brain glucose consumption through the glycolytic pathway in mouse 23 . In Cln7 Δex2 -mCAT LoxP neurons, both glycolysis (Fig. 3a), and its end-product lactate (Fig. 3b) were upregulated (by~1.34 and~1.64-fold, respectively), effects that were abolished in Cln7 Δex2 -CAMKIIa Cre -mCAT neurons (Fig. 3a,  b). Glycolytic and pentose-phosphate pathway (PPP) fluxes are inversely regulated in neurons [27][28][29] . Agreeingly, the increased glycolytic flux observed in primary neurons obtained from Cln7 Δex2 mice was accompanied by reduced PPP flux to a similar Fig. 1 Failure in autophagy causes accumulation of structural and functionally impaired mitochondria in Cln7 Δex2 mouse. a SCMAS/LAMP1 and HSP60/ SCMAS colocalization confocal analyses in primary neurons. DAPI reveals nuclei. Scale bar, 20 µm. b LC3-II western blot analysis in primary neurons incubated with lysosomal inhibitors leupeptin (100 µM) plus NH 4 Cl (20 mM) (Lys. Inh.) for 1 h (ß-actin, loading control). c HSP60 and SCMAS western blot analysis in primary neurons incubated with lysosomal inhibitors leupeptin (100 µM) plus NH 4 Cl (20 mM) (Lys. Inh.) for 1 h (ß-actin, loading control). d OCR analysis (left) and calculated parameters (right) in primary neurons. Data are mean ± SEM from n = 3 independent experiments. e Free complex I (CI) and CI-containing supercomplexes (SC) analyses in primary neurons by BNGE in-gel activity (IGA-CI) and by immunoblotted PVDF membranes against CI subunit NDUFS1. Data are mean ± SEM from n = 3 independent experiments. f Mitochondrial ROS analysis in primary neurons. Data are mean ± SEM from n = 9 (WT), n = 10 (Cln7 Δex2 ) independent experiments. g Representative electron microscopy images and analyses of mouse brain cortex mitochondria. Data are in box plots (the box extends from the 25th to 75th percentiles, the horizontal line indicates the median, and the whiskers go down to the smallest value and up to the largest) from n ≥ 27 mitochondria per condition. Scale bar, 500 nm. (M mitochondria, L lysosome, P peroxisomes). h Free CI and CI-containing SC analyses of mouse brain cortex by BNGE IGA-CI and by immunoblotted PVDF membranes against NDUFS1. Data are mean ± SEM from n = 3 or n = 4 (Cln7 Δex2 ) 8-month old animals. i Mitochondrial ROS analysis in freshly isolated mouse brain cortex neurons. Data are mean ± SEM from n = 3 or n = 4 (WT) animals of 6-month old. Statistical analyses were performed by two-tailed Student's t test. Representative images and western blots out of n ≥ 3 experiments are shown. See also Supplementary Fig. 1 (Fig. 3c). In vivo 2-[ 18 F]fluoro-2-deoxy-D-glucose ([ 18 F] FDG) uptake was unchanged in all analyzed brain areas of the Cln7 Δex2 mouse, according to positron-emission tomography (PET) assessment ( Supplementary Fig. 3a). However, in vivo 1 H-magnetic resonance spectroscopy ([ 1 H]MRS) analysis of the Cln7 Δex2 mouse brain revealed a twofold increase in the concentration of glycine ( Supplementary Fig. 3b, c). Whilst the biosynthesis of glycine via the phosphorylated pathway requires glycolysis 30 , its concentration is not a direct evidence of the glycolytic flux. Therefore, using [ 18 F]FDG-PET and [ 1 H]MRS, Fig. 2 Increased generation of mitochondrial ROS by neurons accounts for impaired mitochondrial accumulation and hallmarks of CLN7 disease in Cln7 Δex2 mouse in vivo. a Mitochondrial ROS analysis in primary neurons from the designed genotype. Data are mean ± SEM from n = 3 or n = 5 (Cln7 Δex2 -mCAT LoxP ) independent experiments. b Representative electron microscopy images of the mouse brain cortex displaying the cristae profile plot of intensities over the maximal axis of the magnified shown mitochondrion (left) and the analyses of mitochondrial area and length (right). Data are in box plots (the box extends from the 25th to 75th percentiles, the horizontal line indicates the median, and the whiskers go down to the smallest value and up to the largest) from n ≥ 136 mitochondria per condition of 3-month-old mice. Scale bars, 600 nm. (M mitochondria, L lysosome, ER endoplasmic reticulum). c Representative images of SCMAS, lipofuscin, GFAP and IBA-1 immunohistochemical analysis of the mouse brain cortex. Data are mean ± SEM from n = 4 animals of 3-month old (three serial slices per mouse). Scale bar, 100 µm. Statistical analyses were performed by one-way ANOVA followed by Tukey's post hoc test. See also Supplementary Fig. 2. Source data are provided as a Source Data file. being approaches that lack cell-level resolution, failed to unambiguously ascertain in vivo upregulation of neuronal glycolysis in Cln7 Δex2 mice. The increased glycolytic flux observed in primary neurons obtained from Cln7 Δex2 mice can be indicative of hyperactive 6-phosphofructo-1-kinase (PFK1) 31,32 , a rate-limiting step of glycolysis that is regulated by fructose-2,6-bisphosphate (F-2,6-P 2 ), a robust positive effector of PFK1 33 . The rate of F-2,6-P 2 formation was enhanced by~1.27-fold in Cln7 Δex2 neurons (Fig. 3d), a result that is compatible with higher activity of 6phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 (PFKFB3) -i.e., the only F-2,6-P 2 -forming isoenzyme found in neurons upon stress conditions 27 . PFKFB3 protein was increased both in primary neurons (~1.51-fold) and in vivo brain cortex (~1.85fold) and cerebellum (~1.41-fold) ( Fig. 3e and Supplementary  Fig. 3d) of the Cln7 Δex2 mice. To elucidate whether in vivo PFKFB3 brain accumulation has neuronal or glial origin, we acutely separated these cell-type groups from the mouse cerebellum using an immunomagnetic approach. As shown in Fig. 3f and Supplementary Fig. 3e, PFKFB3 protein was found enhanced by~4.77-fold in neurons and unaffected in the glia. Since PFKFB3 mRNA abundance was unaltered in Cln7 Δex2 neurons (Fig. 3g), we conjectured that increased PFKFB3 protein could be the consequence of inactivating its degrading pathway 27 . Cln7 Δex2 neurons showed hyperphosphorylation of the anaphasepromoting complex/cyclosome (APC/C) activator protein, Cdh1 ( Fig. 3h and Supplementary Fig. 3f), which is sufficient to inhibit APC/C E3-ligase activity that targets PFKFB3 for proteasomal degradation 27 . To pursue this possibility, we noted that the Ca 2+buffering capacity of bioenergetically compromised mitochondria is impaired 34 (Fig. 3i), an activator of calpaina proteolytic enzyme essential in the signaling cascade leading to Cdh1 hyperphosphorylation 35 . Ca 2+ sequestration reduced both PFKFB3 protein ( Fig. 3j and Supplementary Fig. 3g) and glycolysis ( Fig. 3k) in Cln7 Δex2 neurons, confirming Ca 2+ involvement in increasing glycolytic flux. Ca 2+ -mediated calpain activation proteolytically cleaves p35 into p25-a cofactor of the cyclindependent kinase-5 (Cdk5) 36 that phosphorylates Cdh1 35 . We found an increased p35 cleavage into p25 in Cln7 Δex2 primary neurons and in vivo brain cortex and cerebellum ( Fig. 3l and Supplementary Fig. 3h,i). Inhibition of calpain using the specific inhibitor 36 MDL-28170 rescued p35 cleavage and PFKFB3 increase ( Fig. 3m and Supplementary Fig. 3j). Given that these effects suggest the involvement of Cdk5, Cdk5 was knocked down in Cln7 Δex2 neurons, an action that prevented PFKFB3 increase ( Fig. 3n and Supplementary Fig. 3k). Together, these results indicate the occurrence of a Ca 2+ /calpain-mediated activation of Cdk5/p25 pathway that phosphorylates APC/C-cofactor Cdh1, eventually leading to the stabilization of glycolytic enzyme PFKFB3 in CLN7 disease.
Pharmacological inhibition of PFKFB3 restores mitochondrial alterations and hallmarks of Cln7 Δex2 disease in vivo. In neurons, PFKFB3 destabilization boosts glucose consumption through PPP 27 and prevents damage-associated redox stress 27,37,38 given its role at supplying NADPH(H + )-an essential cofactor of glutathione regeneration 39,40 . We therefore sought to assess whether PFKFB3 activity is related to CLN7 disease. We undertook this by inhibiting PFKFB3 activity using the highly selective, rationally designed 41 compound AZ67. Incubation of Cln7 Δex2 neurons with AZ67 at a concentration that inhibits PFKFB3 activity without compromising survival 42 , prevented the increase in F-2,6-P 2 (Fig. 4a) and glycolysis (Fig. 4b) without affecting mROS ( Fig. 4c and Supplementary Fig. 4a). Interestingly, AZ67 protected Cln7 Δex2 neurons from activation of pro-apoptotic caspase-3 ( Fig. 4d and Supplementary Fig. 4b), suggesting its potential therapeutic benefit. To test this in vivo, AZ67 was intracerebroventricularly administered in Cln7 Δex2 mice daily for 2 months at a dose previously selected according to pharmacokinetic and safety parameters ( Supplementary Fig. 4c, d, e). Electron microscopy analysis revealed that AZ67 did not affect the length or area of brain mitochondria in Cln7 Δex2 mice (Fig. 4e), but it prevented the cristae profile amplitude reduction (Fig. 4e); this may indicate, as observed in other paradigms 24,25 , adaptation of the mitochondrial ultrastructure to a bioenergetically efficient configuration upon glycolysis inhibition. Incubation of Cln7 Δex2 neurons with AZ67 partially restored the impairment in basal respiration (Fig. 4f), indicating a functional improvement of the mitochondria. In vivo, AZ67 prevented the accumulation of SCMAS, lipofuscin, and reactive astroglia in the cortex (Fig. 5a, b), and SCMAS and lipofuscin in the hippocampus and cerebellum ( Supplementary Fig. 5a-c) of the Cln7 Δex2 mice. Hindlimb paralysis 12 in Cln7 Δex2 mice was prevented by AZ67 (Supplementary Movies 1-4), indicating functional recovery. Finally, to assess the possible translational implications of these results, neural precursors cells (NPCs) generated from induced pluripotent cells (iPCs) derived from control and two CLN7 disease patients homozygous for missense mutations ( Fig. 6a-d) were analyzed. CLN7 patients-derived NPCs showed increased SCMAS staining (Fig. 6e) and mROS (Fig. 6f). Furthermore, these cells exhibited condensation of mitochondria in the perinuclear region ( Fig. 6g), an effect that was rectified by AZ67 (Fig. 6h).

Discussion
Here, we found that impaired autophagy pathway in CLN7 disease causes accumulation of dysfunctional mitochondria. These mitochondria exhibit complex I disassembly from supercomplexes, which accounts 19 for the high mROS production that contributes to CLN7 disease pathogenesis, according to the accumulation of SCMAS, lipofuscin, and astrogliosis. The signaling cascade involves a Ca 2+ -mediated, calpain-promoted p25 formation, from p35 cleavage, that activated Cdk5. Active Cdk5 phosphorylates -and inactivates 35 -the E3-ligase APC/C-cofactor Cdh1, which leads 27 to PFKFB3 protein stabilization. Interestingly, Cdk5 also phosphorylates -and inhibits-collapsin response mediator protein 2 (CRMP2) 43 , a cytoskeletal protein found reduced in the CLN6 nclf mutant mouse model of Batten disease 44,45 , suggesting a possible common mechanism in both NCLs. Increased PFKFB3 protein and activity stimulated, in primary neurons obtained from Cln7 Δex2 mice, the flux of glycolysis, a pathway that in healthy neurons is downmodulated to facilitate glucose consumption through the antioxidant PPP pathway 27 . Pharmacological inhibition of PFKFB3 using AZ67 hampered the aberrant increase in neuronal glycolysis and alleviated the hallmarks of CLN7 disease pathogenesis after chronic in vivo intracerebroventricular administration. In contrast to neurons, astrocytes abundantly express PFKFB3 32 , which in part explains the normal high glycolytic phenotype of these cells 27,32 . Fig. 3 Upregulation of PFKFB3 protein and activity via a Ca 2+ /calpain/Cdk5 pathway sustains a high glycolytic flux in Cln7 Δex2 neurons. a Glycolytic flux in primary neurons. Data are mean ± SEM from n = 4 (mCAT LoxP , Cln7 Δex2 -mCAT LoxP ), n = 6 (CaMKIIa Cre -mCAT) or n = 5 (Cln7 Δex2 -CaMKIIa Cre -mCAT) independent experiments. b Lactate released by primary neurons (n = 7-8). Data are mean ± SEM from n = 7 (CaMKIIa Cre -mCAT) or n = 8 independent experiments. c PPP flux in primary neurons. Data are mean ± SEM from n = 5 (WT) or n = 4 (Cln7 Δex2 ) independent experiments. d Rate of P-2,6-P 2 formation in primary neurons. Data are mean ± SEM from n = 3 independent experiments. e Representative PFKFB3 western blot analysis in primary neurons and brain cortex (ß-actin, loading control) and the densitometric quantification of the bands (including the replicas). Data are mean ± SD from n = 6 (WT), n = 7 (Cln7 Δex2 ) independent experiments, or n = 3 animals. f Representative western blots showing PFKFB3 protein abundances in immunomagnetically isolated neurons or glial cells (ß-tubulin III and glial-fibrillary acidic protein or GFAP, loading control for neurons and astrocytes, respectively). g PFKFB3 mRNA analysis by RT-qPCR in primary neurons. Data are mean ± SEM from n = 4 independent experiments (values normalized versus ß-actin). h Representative Cdh1 western blot analysis after PhosTag acrylamide electrophoresis in primary neurons (P-Cdh1, hyperphosphorylated Cdh1; ß-actin, loading control). i Cytosolic Ca 2+ analysis in primary neurons. Data are mean ± SEM from n = 3 independent experiments. j, k Representative PFKFB3 western blot (j) and glycolytic flux (k) analyses in primary neurons incubated with Ca 2+ quelator BAPTA (10 µM; 1 h) (ß-actin, loading control). Data are mean ± SEM from n = 5 (WT), n = 4 (Cln7 Δex2 ) independent experiments. l Representative p35 western blot revealing p35 and its cleavage product p25 in primary neurons and brain cortex (ß-actin, loading control). m Representative p35 and PFKFB3 western blot analyses in primary neurons incubated with calpain inhibitor MDL-28170 (MDL) (100 µM; 24 h) (ß-actin, loading control). n Representative Cdk5 and PFKFB3 western blot analyses in primary neurons transfected with Cdk5 siRNA (siCdk5) or scrambled siRNA (-) (9 nM; 3 days) (ß-actin, loading control). Statistical analyses performed by one-way ANOVA followed by DMS's (a) or Tukey's (b, k) post hoc tests or two-tailed Student's t test (c, d, e, g, i). See also Supplementary  Fig. 3. Source data are provided as a Source Data file.
At the dose of AZ67 administered, we show that glycolysis is inhibited in neurons, but unaltered in astrocytes 42 hence indicating that the main in vivo PFKFB3 target is neuronal. According to the partial recovery of respiration, and to the reduction in the cristae profile amplitude of mitochondria in the PFKFB3-inhibited Cln7 Δex2 neurons, the protection exerted by PFKFB3 inhibition represents an adaptation of mitochondrial shape to a more bioenergetically efficient configurations 24,25 . Abnormal accumulation of mitochondria has also been reported in several forms of lysosomal-storage diseases 46 , although their functional characterization is missing and the impact on other Batten disease pathogenesis unknown. Notably, mitochondrial membranes are required for autophagosomal biogenesis 47 , an observation that opens the possibility that dysfunctional mitochondria may be a contributing factor in autophagy failure in CLN7 disease. In this context, it would be interesting to ascertain whether the bioenergetic alterations herein described in CLN7 disease are shared with other NCLs. If so, pharmacological inhibition of PFKFB3 would be a suitable therapeutic approach worth testing to delay and/or palliate the devastating consequences of each type of currently intractable 48 Batten disease. , and a light-dark cycle was maintained for 12 h. The humidity was 45-65%, and the temperature was 20-25°C. Animals were fed ad libitum with a standard solid diet (17% proteins, 3% lipids, 58.7% carbohydrates, 4.3% cellulose, 5% minerals, and 12% humidity) and given free access to water. Cln7 knockout mouse carrying the European Conditional Mouse Mutagenesis (EUCOMM) tm1d allele by Cre-mediated recombination of the floxed exon 2 of the murine Cln7/Mfsd8 gene (Cln7 Δex2 ) 12 were used. To abrogate mitochondrial ROS selectively in neurons in the Cln7 Δex2 mice in vivo, we crossed Cln7 Δex2 mice with transgenic mice harboring the full-length cDNA encoding catalase fused to the cytochrome c oxidase subunit VIII-mitochondrial leading sequence (mitoCatalase or mCAT), which has incorporated a floxed transcriptional STOP cassette between the mitochondrial-tagged catalase cDNA and the CAG promoter, which were previously generated in our laboratory by homologous recombination in the Rosa26 locus under a C57BL/6 background (mCAT LoxP /+) in order to achieve tissue-and time-specific expression of mCAT in vivo 23 . mCAT LoxP /+ mice were mated with mice harboring Cre recombinase under control of the neuronal-specific CAMKII promoter (CAMKIIa Cre ). The progeny, namely CAMKIIa Cre /+; mCAT LoxP /+, were crossed with Cln7 Δex2 / Cln7 Δex2 mice 12 . The offspring were crossed to obtain the following littermates genotypes: i) +/+; mCAT/+; +/+ (mCAT LoxP ); ii) +/+; mCAT/+; CamKIIa/+ (CAMKIIa Cre -mCAT); iii) Cln7 Δex2 /Cln7 Δex2 ; mCAT/+; +/+ (Cln7 Δex2 -mCA-T LoxP ); iv) Cln7 Δex2 /Cln7 Δex2 ; mCAT/+; CamKIIa/+ (Cln7 Δex2 -CAMKIIa Cre -mCAT).
In vitro Cre recombinase activity induction. Infection with adenovirus carriers of Cre recombinase and empty adenovirus (Control) was used to induce mCAT expression in primary culture of neurons conditionals for mCAT expression (mCat LoxP and Cln7 Δex2 -mCAT LoxP ). The virus, transduced at 10 MOI, was purchased to Gene Transfer Vector Core (University of Iowa). Transduction was performed 3 days before cell recollection, and viral particles were left in the cultures for 24 h.
Induced pluripotent stem cells (iPSC) and neural progenitor cells (NPC) generation. iPSC were generated from the dermal fibroblast of two CLN7 patients (Pa380 and Pa474) (approved by the UCL Research Ethics Committee), then characterized and differentiated to NPC as previously described 50   Freshly purification of neurons from the brain from adult mice. Adult mouse brain (from 6-month-old animals) tissue was dissociated with the Adult Brain Isolation Kit (Miltenyi). Dissociated cells, after removal of debris and red blood cells, neurons were separated with the Neuron Isolation Kit (Miltenyi). The identity of the isolated fraction was confirmed previously 19 by western blot against the neuronal marker microtubule-associated protein 2 (MAP2) and GFAP.
Cell treatments. Neurons in primary culture were incubated with the rationally designed, potent, and highly selective PFKFB3 inhibitor AZ PFKFB3 67 (herein referred as AZ67) 41  Autophagy measurement. To analyze the autophagy pathway, primary neurons were incubated in the absence or presence of the inhibitors of the lysosomal proteolysis leupeptin (100 µM) and ammonium chloride (20 mM) for 1 h. Cells were lysed and immunoblotted against LC3-II to assess autophagy, and against SCMAS and HSP60 to assess mitophagic flux 15 .
Cell transfections. For knockdown experiments, small interfering RNA (siRNA) against CDK5 (siCDK5) (s201147; Thermo Fisher) was used. An siRNA control (siControl) (4390843; Thermo Fisher) was used in parallel. siRNA transfections were performed using the Lipofectamine RNAiMAX reagent (Thermo Fisher) at an siRNA final concentration of 9 nM. A solution of lipofectamine in OptiMEM medium (1:16, vol/vol) was mixed with the siRNAs, previously diluted in OptiMEM (0.2 pmol/µl). This mixture was incubated 5 min at room temperature and then added to cells. Cells were used after 3 days.
Total membrane purification. Membranes were isolated from primary cultures of neurons, or whole-brain homogenates 12 . Cells or tissue were homogenized in sucrose buffer (200 mM sucrose, 50 mM Tris-HCl; pH 7.5, 1 mM EDTA) and centrifuged 5 minutes at 1500 × g. Supernatants were centrifuged at 20,000 × g 10 min. Pellet was then homogenized in extraction buffer (50 mM Tris-HCl; pH 7.5, 1% (vol/vol) Triton-X-100, 1 mM EDTA) and incubated 30 min on ice, followed by a centrifugation at 20,000 × g 10 min. Membranes enriched fraction remained in the supernatant.
Mitochondrial isolation. To obtain the mitochondrial fraction, cell pellets or brain cortex were frozen at −80°C and homogenized (10-12 strokes) in a glass Teflon Potter-Elvehjem homogenizer in buffer A (83 mM sucrose and 10 mM MOPS; pH 7.2). The same volume of buffer B (250 mM sucrose and 30 mM MOPS) was added to the sample, and the homogenate was centrifuged (1000 × g, 5 min) to remove unbroken cells and nuclei. Centrifugation of the supernatant was then performed (12,000 × g, 3 min) to obtain the mitochondrial fraction, which was washed in buffer C (320 mM sucrose; 1 mM EDTA, and 10 mM Tris-HCl; pH 7.4) 19 . Mitochondria were suspended in buffer D (1 M 6-aminohexanoic acid and 50 mM Bis-Tris-HCl, pH 7.0).
Blue-native gel electrophoresis and in-gel activity for complex I. For the assessment of complex I organization, digitonin-solubilized (4 g/g) mitochondria (10-50 μg) were loaded in NativePAGE Novex 3-12% (vol/vol) gels (Life Technologies). After electrophoresis, in-gel NADH dehydrogenase activity was evaluated allowing the identification of individual complex I and complex I-containing supercomplexes bands due to the formation of purple precipitated at the location of complex I 19 . Briefly, gels were incubated in 0.1 M of Tris-HCl buffer (pH 7.4), 1 mg/ml of nitro blue tetrazolium, and 0.14 mM of NADH. Next, a direct electrotransfer was performed followed by immunoblotting against mitochondrial complex I antibody NDUFS1. The direct transfer of BNGE was performed after soaking the gels for 20 min (4°C) in carbonate buffer (10 mM NaHCO 3 ; 3 mM Na 2 CO 3 ·10H 2 O; pH 9.5-10). Proteins transfer to polyvinylidene fluoride (PVDF) membranes was carried out at 300 mA, 60 V, 1 h at 4°C in carbonate buffer. Fructose-2,6-bisphosphate determinations. For F-2,6-P 2 determinations, cells were lysed in 0.1 M NaOH and centrifuged (20,000 × g, 20 min). An aliquot of the homogenate was used for protein determination, and the remaining sample was heated at 80°C (5 min), centrifuged (20,000 × g, 20 min) and the resulting supernatant was used for the determination of F-2,6-P 2 concentrations using a coupled enzymatic reaction 53 . This approach reveals the relative abundance of F-2,6-P 2 generated by PFKFB3 by the coupled enzymatic activities of PFK1 (Sigma) (in the presence of 1 mM fructose-6-phosphate and 0.5 mM pyrophosphate), aldolase (Sigma), and triose-phosphate isomerase/glycerol-3-phosphate dehydrogenase (Sigma). This reaction generates glycerol-3-phosphate and oxidizes NADH (Sigma), producing a reduction in the absorbance at 340 nm that is monitored spectrophotometrically.
Phos-tag SDS-PAGE. For the evaluation of phosphorylation levels of CDH1, primary cultures of neurons were homogenized in extraction buffer (100 mM NaCl, 50 mM Tris pH 8, 1% (vol/vol) NP40). Electrophoresis was performed in 8% (vol/vol) SDS-PAGE gels in the presence of 37.5 µM of PhosTag Acrylamide (ALL-107M, Wako) and 75 µM of MnCl 2 . After electrophoresis, gels were washed three times in transfer buffer with 1 mM of EDTA, before electroblotting.  Supplementary Fig. 6a.
Mitochondrial membrane potential. The mitochondrial membrane potential (Δψ m ) was assessed with MitoProbe DiIC 1 (5) (Life Technologies) (50 nM) by flow cytometry (FACScalibur flow cytometer, BD Biosciences) and expressed in arbitrary units. For this purpose, cell suspensions were incubated with the probe 30 min at 37°C in PBS. Δψ m are obtained after subtraction of the potential value determined in the presence of carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone (CCCP) (10 µM, 15 min) for each sample. A schematic representation of the gating strategy can be found in Supplementary Fig. 6b.
Cytosolic Ca 2+ determination using Fura-2 fluorescence. To estimate the intracellular Ca 2+ levels in neurons we used the fluorescent probe Fura-2 (acetoxymethyl-derivative; Life Technologies) 54 . Neurons were incubated with Fura-2 (2 μM) for 40 min in neurobasal medium at 37°C. Then, cells were washed and further incubated with standard buffer (140 mM NaCl, 2.5 mM KCl, 15 mM Tris-HCl, 5 mM D-glucose, 1.2 mM Na 2 HPO 4 , 1 mM MgSO 4 and 1 mM CaCl 2 , pH 7.4) for 30 min and 37°C. Finally, the standard buffer was removed and experimental buffer (140 mM NaCl, 2.5 mM KCl, 15 mM Tris-HCl, D-glucose, 1.2 mM Na 2 HPO 4 , and 2 mM CaCl 2 , pH 7.4) was added. Emissions at 510 nm, after excitations at 335 and 363 nm, respectively, were recorded in a Varioskan Flash (Thermo) spectrofluorometer at 37°C. Ca 2+ levels were estimated by representing the ratio of fluorescence emitted at 510 nm obtained after excitation at 335 nm divided by that at 363 nm (F335/F363). Background subtraction was accomplished from emission values obtained in Fura-2-lacking neurons. At least, 6 wells were recorded per condition in each experiment and the averaged values are shown, normalized per mg of protein present in the sample.
Bioenergetics. Oxygen consumption rates of neurons were measured in real-time in an XFe24 Extracellular Flux Analyzer (Seahorse Bioscience; Seahorse Wave Desktop software 2.6.1.56). The instrument measures the extracellular flux changes of oxygen in the medium surrounding the cells seeded in XFe24-well plates. The assay was performed on day 7 after cell plating/culture. Regular cell medium was then removed, and cells were washed twice with DMEM running medium (XF assay modified supplemented with 5 mM glucose, 2 mM L-glutamine, 1 mM sodium pyruvate, 5 mM HEPES, pH 7.4) and incubated at 37°C without CO 2 for 30 min to allow cells to pre-equilibrate with the assay medium. Oligomycin, FCCP or antimycin/rotenone diluted in DMEM running medium were loaded into port-A, port-B, or port-C, respectively. Final concentrations in XFe24 cell culture microplates were 1 μM oligomycin, 2 μM FCCP, 2.5 μM antimycin, and 1.25 μM rotenone. The sequence of measurements was as follow unless otherwise described. The basal level of oxygen consumption rate (OCR) was measured three times, and then port-A was injected and mixed for 3 min after OCR was measured three times for 3 min. Same protocol with port-B and port-C. OCR was measured after each injection to determine the mitochondrial or non-mitochondrial contribution to OCR. All measurements were normalized to average three measurements of the basal (starting) level of cellular OCR of each well. Each sample was measured in three to five wells. Experiments were repeated three to five times with different cell preps. Non-mitochondrial OCR was determined by OCR after antimycin/rotenone injection. Maximal respiration was determined by maximum OCR rate after FCCP injection minus non-mitochondrial OCR. ATP production was determined by the last OCR measurement before oligomycin injection minus the minimum OCR measurement after oligomycin injection.
Activity of mitochondrial complexes. Cells were collected and suspended in PBS (pH 7.0). After three cycles of freeze/thawing, to ensure cellular disruption, complex I, complex II, complex II-III, complex IV, and citrate synthase activities were determined. Rotenone-sensitive NADH-ubiquinone oxidoreductase activity (complex I) 55  Protein determinations. Protein samples were quantified by the BCA protein assay kit (Thermo) using BSA as a standard.
Pharmacokinetics of AZ67. For the pharmacokinetic assay, healthy male C57BL/6 mice were used. A single dose of 40 mg/kg of AZ67 was injected intravenously and the blood, cerebrospinal fluid (CSF) and brain, were collected after 5 min, 15  In vivo toxicity assay. Male mice (C57BL6/J; six animals per group; 8-week old) (purchased from Charles-River, Spain) were subjected to the implantation of a cannula in the lateral ventricle under anesthesia and then left for at least 15 days for full recovery. After this, the PFKFB3 inhibitor (AZ67) was administered through the cannula using an automatic micro-pump (CMA 4004 Microdialysis Syringe Pump, Harvard Apparatus) at different doses: 0 (vehicle), 0.005, 0.01, 0.05, 0.1, 1, and 10 nmol/mice. The compounds were administered every 24 h for 1 week, and animals were analyzed in the open field immediately before each administration. We selected the maximal dose that caused no evident alterations and/or deterioration of the animals for the following experiments, being 1 nmol/mouse.
Open-field tests. Male mice were left to acclimate in the room for no less than 15 min at the same time of day (10:00 to 14:00). Tracking was carried out one at a time, and we carefully cleaned the apparatus with 70% ethanol between trials to remove any odor cues. An ANY-box core was used, which contained a light-gray base and an adjustable perpendicular stick holding a camera and an infrared photobeam array to track the animal movement and to detect rearing behavior, respectively. Mouse movements were tracked with the ANY-maze 5.33 software and the ANY-maze interface to register all parameters described subsequently. For the open-field test, a 40 × 40 × 35 cm (w, d, h), black infrared transparent Perspex insert was used, and the arena was divided into three zones, namely border (8 cm wide), center (16% of the total arena) and intermediate (the remaining area). The test lasted for 10 min, and the distance traveled, and the time spent in each zone was measured.
AZ67 in vivo administration. AZ67 (Tocris) for in vivo usage was dissolved in 20% (wt/vol) PEG200 in PBS to a 20 mM concentration. Four groups were generated (four to six animals/group), namely: WT-vehicle, Cln7 Δex2 -vehicle, WT-AZ67, Cln7 Δex2 -AZ67. The cannula was inserted intracerebroventricularly at the age of 8 weeks and, after at least 15 days of recovery, we injected the AZ67 at the dose identified previously (1 nmol/mouse) every 24 h. The duration of the experiment was determined by the presence of hindlimb clasping the Cln7 Δex2 vehicle-treated mice, being this time two months. After this, the animals were perfused, and their brains dissected to be investigated by immunofluorescence and electron microscopy.
Electron microscopy and mitochondrial morphology analysis. Male mice were anaesthetized by intraperitoneal injection of a mixture of xylazine hydrochloride (Rompun; Bayer) and ketamine hydrochloride/chlorbutol (Imalgene; Merial) (1:4) at 1 ml per kg body weight and then perfused intra-aortically with 0.9% NaCl followed by 5 ml/g body weight of 2% (wt/vol) paraformaldehyde plus 2% (vol/vol) glutaraldehyde. After perfusion, brains were dissected out sagitally in two parts and post-fixed with perfusion solution overnight at 4°C. Brain blocks were rinsed with 0.1 M PB solution and a 1 mm 3 squared of brain cortex was excised and treated with osmium tetroxide (1% in PB) for 1 h. Tissue was then washed with distilled water and dehydrated in ascending series of ethanol followed by embedment in EPON resin. Ultra-thin sections (50 nm) were stained with uranyl acetate and lead citrate and examined with Tecnai Spirit Twin 120 kv transmission electron microscopy equipped with a digital camera Orius WD or JEM-1010 (JEOL) 100 kv transmission electron microscopy equipped with a digital camera AMT RX80. For mitochondrial area quantification, the area of each mitochondrion was quantified in neuronal soma, axons and dendrites. In the case of mitochondrial length, the values represent the length in the maximal axis of mitochondria in the plane of microphotographies. Cristae profiles of representative mitochondria of each condition and type were traced along the major axis that crosses mitochondria perpendicularly to cristae. Data of pixel intensity were obtained using the plot profile plugin of ImageJ software.
Imaging and quantification. Sections were examined with epifluorescence and the appropriate filter sets under an Operetta CLS high-content imaging system (Per-kinElmer). Large fields of view were acquired with an ×5 scan using an OperaPHX/ OPRTCLS ×5 Air Objective. Then high-resolution images were acquired using an OperaPHX/OPRTCLS Air Objective ×20 hNA objective. Immunohistochemical digital images were used to analyze different protein staining in the three most sagittal sections per animal. Images were analyzed with the Harmony software with PhenoLOGIC (PerkinElmer). Interest brain areas (cortex, hippocampus, and cerebellum) were selected and subsequently quantified as mean intensity per area by using the "measure rectangle" function, which represents the mean intensity of a channel per selected area.
NPC immunocytochemistry. NPCs were fixed with 100% ice-cold methanol for 5 min and incubated in blocking solution (1% (vol/vol) normal goat serum, 0.1% (wt/vol) bovine serum albumin (BSA), 0.1% (vol/vol) Triton X-100 in DPBS). The antibodies were incubated in a blocking solution. The incubation of the primary antibody (anti-ATP5A, (1/100) (ab14748; Abcam) or SCMAS (1/200) (ab181243; Abcam) was performed for 2 h at room temperature, and the secondary antibodies (Alexa Fluor 568 goat anti-mouse or Alexa Fluor 488 goat anti-rabbit (1/500)) were applied for 1 h at room temperature. Slides were mounted with VECTASHIELD Mounting Medium with DAPI, incubated for 24 h at 4°C, and imaged with a Zeiss Axio Imager M2 fluorescence microscope or under an inverted microscope (Nikon; Eclipse Ti-E) equipped with a pre-centered fiber illuminator (Nikon; Intensilight C-HGFI), B/W CCD digital camera (Hamamatsu; ORCA-E.R.). Fluorescence quantification was performed after appropriate thresholding using the ImageJ software (NIH). The pixel intensity profile of ATP5A immunodecoration was analyzed across the maximal axis of the cell that departs from the nucleus, using the plot profile plugin of ImageJ software. A representative profile is shown for each condition.  (FLUORSCAN 3000). Before imaging studies, animals (5-month old) were fastened for 6 h with free access to drinking water. Administration of [ 18 F]FDG (19.2 ± 1.6 MBq, 100 µL) was carried out via one of the lateral tail veins under anesthesia, induced with 3.0-5.0% isoflurane in pure oxygen and maintained with 1.5-2.0% isoflurane in pure oxygen. After administration, animals were allowed to recover from anesthesia for 45 minutes before being subjected to positron-emission tomography (PET) studies. PET studies (n = 5 for control and study groups; 10 min acquisitions) were conducted using the β-cube microsystem (Molecubes, Gent, Belgium), with the head of the animal positioned in the center of the field of view, in one-bed position using a 511 keV ± 30% energetic window. A computerized tomography (CT) scan was acquired immediately after the finalization of the PET imaging session, both for anatomical reference and to determine the attenuation map for PET image reconstruction. PET images were reconstructed with OSEM-3D iterative algorithm. Images were analyzed using π-MOD image analysis software (π-MOD Technologies Ltd, Zurich, Switzerland). With that aim, PET images were manually coregistered with a M. Mirrione-T2 MRI atlas available at π-MOD software. Volumes of interest (VOIs) were automatically delineated in different brain regions, namely cortex, cerebellum, brain stem, hippocampus, striatum, and whole brain, and the concentration of radioactivity in each region was determined and decay-corrected to injection time. Values were finally normalized to injected amount of radioactivity and body weight, to obtain standard uptake values (SUVs).
Magnetic resonance spectroscopy. Localized 1 H-MRS was performed at 11.7 Tesla using a 117/16 USR Bruker Biospec system (Bruker Biospin GmbH, Ettlinglen, Germany) interfaced to an advance III console and operating ParaVicion 6.1 under topspin software (Bruker Biospin). After fine-tuning and shimming of the system, water-signal FWHM values typically in the 15-25 Hz range were achieved. Scanning started with the acquisition of three scout images (one coronal, one transverse, and one sagittal) using a 2D-multiplane T2W RARE pulse sequence with Bruker's default parameters. Those images were used to place the spectroscopy voxel of size 1.5 × 1.5 × 2 mm 3 located at the right striatum of the mouse brain or 2 × 0.8 × 2 mm 3 located in the cortex (at the mid-line of the brain), always with care not to include the ventricles in the voxel (the geometry of the voxel was slightly altered to avoid this event, when necessary). At least two 1 H-MRS spectra were acquired per scanning session per animal (5-month-old animals). The voxel was repositioned, and shimming adjustments were repeated between acquired spectra, when the spectral resolution of the obtained 1 H-spectrum was not good. For 1 H-MR a water suppressed PRESS sequence was used with the following parameters: Echo time = 17.336 ms (TE1 = TE2 = 8.668 ms); Repetition time = 2500 ms; Naverages = 256; Acquisition size = 2048 points; spectral width = 11 ppm (5498.53 Hz). MR spectra were fitted and quantified using LC-Model 6.3-1R 63 .
Statistical analysis. The comparisons between two groups of values we performed using two-tailed Student's t test. For multiple-values comparisons, we used oneway ANOVA followed by either Tukey or DMS post hoc tests, as indicated in the figure legends. The statistical analysis was performed using the GraphPad Prism v8 software. The number of biologically independent culture preparations or animals used per experiment are indicated in the figure legends, and the P values in the figures.