Altered neuronal lactate dehydrogenase A expression affects cognition in a sex- and age-dependent manner

Summary

The astrocyte-neuron lactate shuttle (ANLS) model posits that astrocyte-generated lactate is transported to neurons to fuel memory processes. However, neurons express high levels of lactate dehydrogenase A (LDHA), the rate-limiting enzyme of lactate production, suggesting a cognitive role for neuronally generated lactate. It was hypothesized that lactate metabolism in neurons is critical for learning and memory. Here transgenic mice were generated to conditionally induce or knockout (KO) the Ldha gene in CNS neurons of adult mice. High pattern separation memory was enhanced by neuronal Ldha induction in young females, and by neuronal Ldha KO in aged females. In older mice, Ldha induction caused cognitive deficits whereas Ldha KO caused cognitive improvements. Genotype-associated cognitive changes were often only observed in one sex or oppositely in males and females. Thus, neuronal-generated lactate has sex-specific cognitive effects, is largely indispensable at young age, and may be detrimental to learning and memory with aging.

Graphical abstract

Highlights

• •Altered neuronal Ldha expression elicits cognitive changes depending on sex and age
• •Induced neuronal Ldha causes cognitive deficits in old mice
• •Knockout of neuronal Ldha causes cognitive improvements in old mice
• •Neuronal lactate production may help or hinder cognition depending on the context

Neuroscience; Behavioral neuroscience; Molecular neuroscience; Cellular neuroscience; Sensory neuroscience

Introduction

Mammalian brains require a disproportionate amount of energy compared to the rest of the body.^1,2,3,4,5 These energetic needs are met mainly through carbohydrate metabolism^6,7,8 that differs between neurons and astrocytes, the two principle brain cell types. Astrocytes predominately break down sugars into lactate using glycolysis or glycogenolysis.^9,10,11 Neurons can also utilize glycolysis, but generally favor oxidative metabolism involving further break down of glycolytic end-products through the tricarboxylic acid (TCA) cycle and mitochondrial oxidative phosphorylation.^12,13,14 Astrocytes support neuronal metabolism by releasing lactate that can be taken up by neurons and used for oxidative metabolism through a process known as the astrocyte-neuron lactate shuttle (ANLS).^15,16 While numerous studies have shown that the ANLS plays a key role in learning and memory,^{17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53} these findings were based on experiments in which lactate transport or glycogenolysis were inhibited using a pharmaceutical approach with potentially non-cell-type specific actions or using a genetic approach targeting either astrocytes or non-specific cell types. However, the effect of directly manipulating lactate production within certain CNS cell types on cognition has not been examined before.

There are two predominant lactate dehydrogenase (LDH) isoforms in the mammalian brain, LDHA and LDHB, which form a tetramer in either a homo- or heteromeric fashion resulting in five possible combinations. The number of LDHA or LDHB isoforms that make up an LDH tetramer influences its activity; LDHB predominantly converts lactate to pyruvate whereas LDHA predominantly converts pyruvate to lactate.^{54,55,56,57,58,59,60,61,62,63,64,65} Thus neurons are predicted to express high levels of LDHB in order to oxidatively metabolize lactate to pyruvate, whereas astrocytes are predicted to express high levels of LDHA to reduce pyruvate to lactate.^66,67,68 However, recent studies using cell-type distinction methods have revealed that LDHA is readily detectable in neurons.^{69,70,71,72,73,74,75} Furthermore, neurons can take up glucose directly^76,77 and, under certain contexts, exhibit increased glycolysis and lactate release.^{52,78,79,80,81} Therefore, neurons have the capacity to either break down astrocyte-derived lactate, in accordance with the ANLS, or produce lactate glycolytically in a cell autonomous manner, and this choice may be biased in one direction depending on the levels or activity of LDHA. The question arises as to why neurons express high levels of LDHA and what role neuronal produced lactate plays in cognition? In addition, how neuronal lactate metabolism changes with age or between sexes is unknown.

To address these questions, two separate neuronal transgenic mice were generated with Ldha either induced or knocked out in neurons of the brain. Different cognitive domains were tested in neuronal Ldha transgenic mice using a variety of behavioral assays. In most paradigms used, altered neuronal Ldha expression resulted in either cognitive improvements or deficits that were also age- and sex-dependent. Importantly, in older mice, all cognitive changes affected by neuronal Ldha induction were detrimental, whereas cognition was improved by neuronal Ldha knockout (KO). These results highlight that elevated neuronal lactate metabolism elicits sex-dependent effects that may have detrimental consequences on cognition with age.

Results

LDHA is increased in the brains of neuronal Ldha induction mice

To manipulate LDHA in the brains of mice across age, two separate systems were employed to either induce or KO the Ldha gene in central nervous system (CNS) neurons. A novel transgenic mouse line was generated containing a tetracycline-responsive promoter element (TRE) upstream of the mouse Ldha gene fused to a hemagglutinin (HA) tag combined with the Tet-Off system^82,83 to selectively induce HA-tagged LDHA in neurons (Figure 1A). HA-tagged Ldha expression was restricted to excitatory neurons in adulthood by employing the calcium-calmodulin-dependent kinase II (CaMKIIα) driver upstream of the tetracycline transactivator gene to turn off TRE-Ldha expression during embryonic and post-natal development using dietary doxycycline (Figure 1A). The CaMKIIα driver was selected because it strongly induces neuronal expression in the hippocampus and frontal cortex with only minor expression in the cerebellum.^84,85 In neuronal Ldha induction mice of both sexes, HA-tagged LDHA was detectable across age in the hippocampus (Figures 1BI and S1), frontal cortex (Figures 1BII and S1), but not cerebellum (Figures 1BIII and S1) by western blotting, and localized to microtubule-associated protein 2 (MAP2) positive neurons by immunofluorescence analysis (Figures 1C–1F and S2–S7). Interestingly, hippocampal HA-tagged LDHA levels, measured by western blotting, in neuronal Ldha induction mice were higher in 14-month-old mice compared to 7-month-old mice (Figures 1BI and S1).

Figure 1.

Neuronal Ldha induction transgenic mice express HA-tagged LDHA in the brain

(A) Schematic outlining Tet-Off system used to induce expression of Ldha in neurons. Neuronal Ldha induction mice contain both the genetic constructs encoding the tetracycline-controlled transactivator (tTA) and the tetracycline-responsive promoter element (TRE). Control mice lack either one or both constructs. The tTA induces expression of Ldha downstream of TRE following removal of doxycycline from the diet.

(B) Western blot analysis demonstrating HA-tagged LDHA expression in the hippocampus (I) and frontal cortex (II), but not cerebellum (II), of neuronal Ldha induction mice at young (7 months) and old (18 months) age. HA-tagged LDHA increased with age in neuronal Ldha induction mice (I; genotype × age effect: F(1, 20) = 13.54, p = 0.0015). n = 6. Comparisons made by two-way ANOVA, fixed effects presented in each graph, with Šídák multiple comparisons between genotype for each age or, if there was an effect including age, Holm-Šídák multiple comparisons between genotype for each age and between age for each genotype. See also Figure S1.

(C) Representative immunofluorescence images revealed increased levels of HA-tagged LDHA in the hippocampus of neuronal Ldha induction mice at young (7 months) and old (18.5 months) age compared with age-matched control. Scale bar: 500 μm. See also Figures S2 and S5.

(D) Magnified view of immunofluorescence image from hippocampal Cornu Ammonis 2 (CA2) region of 18.5 month old female neuronal Ldha induction mouse revealing neuronal localization of HA-tagged LDHA. Scale bar: 50 μm.

(E and F) Quantification of immunofluorescence staining showing the region-dependent increased percentage of MAP2 and HA-tagged LDHA positive neurons in neuronal Ldha induction mice compared to control mice. Increases were detected in hippocampus (HPC), frontal cortex (CTX), and cerebellum (CBM) at young age (E; brain region effect: F(1.816, 21.80) = 17.26, p < 0.0001; genotype effect: F(1, 12) = 63.20, p < 0.0001; brain region × genotype effect: F(2, 24) = 23.75, p < 0.0001) and old age (F; brain region effect: F(1.167, 16.34) = 28.98, p < 0.0001; genotype effect: F(1, 14) = 40.28, p < 0.0001; brain region × genotype effect: F(2, 28) = 29.47, p < 0.0001). n = 8. See also Figures S2–S7. Comparisons made using a mixed-effects model with Geisser-Greenhouse correction, fixed effects presented in each graph, and Holm-Šídák multiple comparisons between genotypes for each brain region and between each brain region for transgenic. Data presented as mean ± SEM.

LDHA is reduced in the brains of neuronal Ldha knockout mice

The ligand-activated site-specific Cre-lox system^86,87 was used to generate neuronal Ldha KO mice using the same neuronal driver, CaMKIIα, as in the neuronal Ldha induction mice (Figure 2A). Ldha KO was restricted to adulthood by administering tamoxifen at 4–5 months of age. In neuronal Ldha KO mice at young and old ages, western blotting revealed a decrease in LDHA protein levels in the hippocampus (Figures 2BI and S8) and frontal cortex (Figures 2BII and S8), with no effect in the cerebellum (Figures 2BIII and S8). In contrast, LDHB protein levels were unaltered (Figure 2C). Immunofluorescence staining also showed reduced LDHA in the brains of neuronal Ldha KO mice (Figures 2D, 2E, and S9–S11).

Figure 2.

Neuronal Ldha knockout mice exhibit reduced LDHA in the brain

(A) Schematic outlining Cre-lox system used for knocking out Ldha in neurons. Neuronal Ldha KO mice contain both a genetic construct encoding the tamoxifen-dependent cyclization recombination (Cre) recombinase and mutant human estrogen receptor ligand-binding domain chimera (CreERT2) and two locus of crossing-over of bacteriophage P1 (loxP) recognition sites flanking exon 3 of the Ldha gene. Control mice lack CreERT2. CreERT2 permits the excision of exon 3 of Ldha in the presence of tamoxifen.

(B) Western blot analysis demonstrating LDHA protein levels are reduced in the hippocampus (I; genotype effect: F(1, 20) = 6.353, p = 0.0203) and frontal cortex (II; genotype effect: F(1, 20) = 30.24, p < 0.0001), but not cerebellum (II), of neuronal Ldha KO mice at young (6 months) and old (18 months) age. n = 6. See also Figure S8.

(C) Western blot analysis demonstrating LDHB protein levels are not reduced in hippocampus (I), frontal cortex (I), or cerebellum (II) of neuronal Ldha KO mice at young (6 months) and old (18 months) age. n = 6. Comparisons made by two-way ANOVA, fixed effects presented in each graph, with Šídák multiple comparisons between genotype for each age or, if there was an effect including age, Holm-Šídák multiple comparisons between genotype for each age and between age for each genotype. See also Figure S8.

(D) Representative immunofluorescence images showing reduced LDHA in the hippocampus of neuronal Ldha KO mice at young (8 months) and old (18.5 months) age compared with age-matched control. Scale bar: 500 μm. See also Figure S9.

(E) Quantification of immunofluorescence staining showing a decrease in CaMKIIα positive neurons containing LDHA in neuronal Ldha KO mice compared to control mice in the hippocampus (I; genotype effect: F(1, 28) = 13.52, p = 0.0010), frontal cortex (II; genotype effect: F(1, 28) = 10.56, p = 0.0030), and cerebellum (III; genotype effect: F(1, 28) = 9.508, p = 0.0046) at young (8 months) and old (18.5 months) of age. The percent of CaMKIIα and LDHA positive neurons decreased with age in the hippocampus (age effect: F(1, 28) = 8.532, p = 0.0068) and frontal cortex (age effect: F(1, 28) = 27.76, p < 0.0001) in neuronal Ldha KO and control mice. n = 8. See also Figures S9–S11. Comparisons made using a three-way ANOVA, fixed effects presented in each graph, and Holm-Šídák multiple comparisons between genotype within each age and between ages for each genotype. Data presented as mean ± SEM.

Neuronal Ldha induction raises hippocampal lactate levels

To assess whether altered LDHA in neurons causes a change in hippocampal lactate, neuronal Ldha induction and KO mice were analyzed using in vivo ¹H-magnetic resonance spectroscopy (¹H-MRS). Young female neuronal Ldha induction mice exhibited increased hippocampal lactate levels compared to age-matched control mice (Figure 3AI), whereas old neuronal Ldha induction mice did not (Figure 3AII). Neuronal Ldha KO mice did not have any detectable changes in hippocampal lactate at young (Figure 3BI) or old (Figure 3BII) age compared to age-matched control mice when measured by ¹H-MRS. In addition, metabolite levels in the hippocampus of neuronal Ldha induction and KO mice were analyzed using gas chromatography-mass spectrometry (GC-MS). Notably, perfusion during extraction of the hippocampus for GC-MS analysis allows for measurement of intracellular lactate without confounding cerebrovascular lactate present in the brain during in vivo ¹H-MRS. GC-MS analysis revealed increased hippocampal lactate levels in both young and old age neuronal induction mice compared with control mice (Figure 3C). In contrast, increased hippocampal lactate was detected with age, regardless of genotype, in the hippocampus of neuronal Ldha KO and age-matched control mice (Figure 3D). Including all features detected by GC-MS, a principal component analysis (PCA) was conducted to identify outliers (Figures S12 and S13) and partial least squares-discriminant analysis (PLSDA) to identify the metabolites with the highest variable importance in projection (VIP) scores (Figures 3E and 3F).

Figure 3.

Hippocampal lactate levels are elevated in neuronal Ldha induction mice

(A) In vivo¹H-MRS analysis revealed elevated hippocampal lactate levels in young (8 months) neuronal Ldha induction female mice compared to age-matched control mice (I; n = 8; t(14) = 2.611, p = 0.0205) whereas no genotype effects on lactate levels were detected in old (15 months) mice (II; n = 8).

(B) In vivo¹H-MRS analysis of lactate levels revealed similar hippocampal lactate levels when comparing neuronal Ldha KO to age-matched control mice at both young age (I; n = 7; 9.5 months) and old age (15 months). Comparisons for ¹H-MRS analysis were made using unpaired t test if only one sex measured and using two-way ANOVA, fixed effects presented in each graph, with unpaired t test between genotypes.

(C) GC-MS analysis revealed elevated hippocampal lactate levels in neuronal Ldha induction mice across both young and old (genotype effect: F(1, 37) = 5.628, p = 0.0230) compared to young and old control.

(D) GC-MS analysis revealed in neuronal Ldha KO mice only elevation of hippocampal lactate with age across both genotypes (age effect: F(1, 37) = 18.61, p = 0.0001; control: t(18) = 3.139, p = 0.022521; KO: t(19) = 2.963, p = 0.023766) with no difference between genotypes. Comparisons for GC-MS analysis were made using two-way ANOVA, fixed effects presented in each graph, with Holm-Šídák multiple comparisons. p values for partial correlation analysis were Bonferonni corrected.

(E and F) Heatmaps depicting relative abundance of the 25 metabolites with the top variable importance in projection (VIP) scores identified by partial least squares-discriminant analysis (PLSDA) of metabolomes generated by GC-MS analysis of the hippocampus in neuronal Ldha induction and age-matched control mice (E) or neuronal Ldha KO and age-matched control mice (F). Metabolites that could not be identified in the NIST database are listed as a number based on their mean retention time. See also Figures S12 and S13. Data presented as mean ± SEM.

Ascorbic acid was elevated in young female neuronal Ldha induction mice relative to age-matched control mice. In contrast, L-aspartic acid and L-glutamic acid were slightly elevated in neuronal Ldha KO mice relative to age-matched control mice.

Neuronal Ldha transgenic mice do not have locomotor abnormalities

Locomotor abnormalities can be a confounding factor when assessing cognitive behavioral tasks in mice. To ensure that Ldha transgenic mice had normal motor function, locomotor ability was assessed on the rotarod (Figure 4A). Neither neuronal Ldha induction (Figure 4B) nor Ldha KO (Figure 4C) mice had deficits in locomotor learning or maximum ability on the rotarod at old age. In fact, old neuronal Ldha KO mice were better able to familiarize themselves to the rotarod (Figure 4CI). Locomotor function was also assessed by measuring swimming speed using the Morris water maze (MWM) (Figure 4D). Relative to control mice, neuronal Ldha induction (Figures 4E–4H) and Ldha KO (Figures 4I–4L) mice did not exhibit any changes in swimming ability at young or old age during training days or probe trials.

Figure 4.

Locomotor ability is not impacted by neuronal Ldha induction or knockout

(A) In the rotarod paradigm mouse locomotor ability was determined by assessing familiarization to a spinning rod on day 1, followed by learning to remain on a spinning rod as it progressively accelerates across each of five consecutive trials on day 2 and with maximum performance assessed on day 2 trial 6.

(B) Locomotor ability on the rotarod for 14.5 month old neuronal Ldha induction and age-matched control mice for either sex did not differ on day 1 (I) or day 2 (II). Transgenic mice did not differ from age-matched control mice for day 1 number of falls (I), day 2 latency to fall (II) or day 2 at maximum ability (II). n = 5–9.

(C) Locomotor ability on rotarod for 14.5 month old Ldha KO and age-matched controls on day 1 familiarization to the rod differed by sex and genotype (I; n = 8–12; sex effect: F(1, 39) = 11.56, p = 0.0016; genotype effect: F(1, 39) = 4.432, p = 0.0418) and on day 2 differed by sex (II; n = 9–12; sex effect trial 1–5: F(1, 40) = 10.61, p = 0.0023; sex effect trial 6: F(1, 40) = 4.988, p = 0.0312). KO and control mice did not differ on day 2 (II). Comparisons of rotarod performance on day 1 and day 2 trial 6 were made using a two-way ANOVA across sex and genotype, fixed effects presented in each graph, and unpaired t test between genotypes. Comparisons for rotarod day 2 trial 1–5 were made using a mixed-effects model with Geisser-Greenhouse correction, fixed effects presented in each graph, and Šídák’s multiple comparisons test between genotype for each trial, within each sex if there was an effect of sex.

(D) Locomotor activity assessed by measuring average swimming speed during training day trials and probe trials in the Morris water maze as depicted in Figure 7A.

(E) Locomotor ability on training days was retained equally in 6 month old neuronal Ldha induction and age-matched control mice for both sexes. n = 9–12.

(F) Locomotor ability in young aged (6 months) was retained equally in neuronal Ldha induction compared to age-matched control mice of both sexes for mean swimming speed in the 24 h probe trial. n = 9–12.

(G) Locomotor ability on training days was retained equally in old aged (12.5 months) neuronal Ldha induction and control mice for both sexes. n = 8–13.

(H) Locomotor ability in old aged (12.5 months) neuronal Ldha induction and age-matched control mice of both sexes was retained equally in the 24 h probe trial. n = 8–13.

(I) Locomotor ability on training days in young aged (7 months) neuronal Ldha KO compared to age-matched control mice was retained equally. n = 12–13.

(J) Locomotor ability in young aged (7 months) neuronal Ldha KO compared to age-matched control mice was retained equally in the 24 h probe trial. n = 12–13.

(K) Locomotor ability on training days in young aged (13 months) neuronal Ldha KO and age-matched control mice of both sexes were retained equally. n = 9–12.

(L) Locomotor ability in old aged (13 months) neuronal Ldha KO and age-matched control mice of both sexes were retained equally in the 24 h probe trial. n = 9–12. For training days mean speed swimming, comparisons were made using a mixed-effects model with Geisser-Greenhouse correction, fixed effects presented in each graph, and further comparisons between genotype for each day were made using unpaired t tests for conditions with an effect including genotype. For probe trials, comparisons were made using a two-way ANOVA, fixed effects presented in each graph, and further comparisons between genotype made using unpaired t tests, within each sex for conditions with an effect including sex for training days or probe trials. Data presented as mean ± SEM.

Neuronal Ldha transgenic mice have altered anxiety-like behavior

Emotional state and temperament can influence cognitive ability.^88,89 To gauge the extent to which temperament is influenced by modifying neuronal Ldha expression, anxiety-like behavior was evaluated using two different paradigms, thigmotaxis (Figure 5A) and the light-dark box (Figure 5D). Thigmotaxis is an anxiety-like behavior defined by the centrophobic tendency to stay near the walls of a novel open space.^90,91 Among young and old neuronal Ldha induction mice, only old induction males exhibited an increase in thigmotaxis anxiety-like behavior compared with age-matched controls (Figure 5B). Thigmotaxis anxiety-like behavior in neuronal Ldha KO mice was significantly lower in young mice but unchanged at old age compared with age-matched controls (Figure 5C). Light-dark box anxiety-like behavior is the propensity to avoid brightly lit spaces. A longer latency to enter the dark side, or decreased time spent in the dark side, is reflective of reduced anxiety. Light-dark box and thigmotaxis in white light both measure mouse behavior analogous to human state anxiety,^92,93 yet testing with anxiogenic and anxiolytic drugs indicate that thigmotaxis is more sensitive to anxiety increases compared to anxiety decreases.⁹⁰ In the light-dark box, young male neuronal Ldha induction mice exhibited significantly reduced anxiety-like behavior compared to age-matched control mice (Figure 5EII). In contrast, young male neuronal Ldha KO mice in the light-dark box exhibited increased anxiety-like behavior compared to control mice (Figure 5FII).

Figure 5.

Thigmotaxis and light-dark box anxiety-like behavior are differentially impacted by neuronal Ldha induction or knockout depending on age and sex

(A) Thigmotaxis is a type of anxiety-like behavior that was measured as time spent in outer zone during the first day of habituation to the circular arena during the spontaneous location recognition paradigm (Figure 6F).

(B) Thigmotaxis anxiety-like behavior in neuronal Ldha induction mice compared to age-matched control mice was unchanged at young age (5 months) and higher at old age (12 months) for males (t(43) = 2.819, p = 0.0144; sex × genotype effect: F(1, 43) = 5.943, p = 0.0190) but not females. n = 8–15.

(C) Thigmotaxis anxiety-like behavior in neuronal Ldha KO mice compared to age-matched control mice was lower at young age (6 months; t(43) = 2.641, p = 0.0115; genotype effect: F(1, 41) = 6.746, p = 0.013) and unchanged at old age (12.5 months). N = 20–24. For thigmotaxis, comparisons made by two-way ANOVA within each age group, fixed effects presented in each graph, with Šídák’s multiple comparisons between genotypes.

(D) Anxiety-like behavior was assessed in the light-dark box as time spent on dark side and latency to enter dark side.

(E) Light-dark box anxiety-like behavior of young age (6.5 months) neuronal Ldha induction mice and age-matched control differed by sex for latency to enter dark side (I; F(1, 33) = 5.601, p = 0.024), and time on dark side (II; F(1, 34) = 13.31, p = 0.0009). Transgenic males had decreased anxiety-like behavior for time on dark side (II; sex × genotype effect: F(1, 34) = 11.10, p = 0.0021; t(34) = 4.186, p = 0.0004). n = 5–13.

(F) Light-dark box anxiety-like behavior of young aged (7 months) neuronal Ldha KO mice and age-matched control differed by sex for time on dark side (II; F(1, 41) = 5.180, p = 0.0281). KO males had increased anxiety-like behavior for time on dark side (II; sex × genotype effect: F(1, 41) = 6.612, p = 0.0139; t(41) = 2.47, p = 0.0352). n = 8–12. Comparisons for light-dark box made using two-way ANOVA, fixed effects presented in each graph, and Šídák’s multiple comparisons between genotype for each sex. Data presented as mean ± SEM.

Neuronal Ldha transgenic mice do not exhibit changes in long-term recognition memory

Recognition memory is a type of declarative memory⁹⁴ that is most reliant on the perirhinal cortex.⁹⁵ The spontaneous object recognition (SOR) test in rodents increasingly depends on the perirhinal cortex dependent recognition memory when conducted in a Y-shaped maze.⁹⁶ SOR testing with a Y-shaped maze was used to measure 24 h long-term recognition memory in young and old neuronal Ldha induction and KO transgenic mice (Figure 6A). Neuronal Ldha induction (Figures 6B and 6C) and Ldha KO mice (Figures 6D and 6E) did not exhibit changes in recognition memory compared to age-matched controls at young or old age. Moreover, in both transgenic lines and their controls, recognition memory was not lost at the older ages (Figures 6C and 6E).

Figure 6.

Short-term spatial memory with high pattern separation is increased by neuronal Ldha induction in young and by neuronal Ldha knockout in old female mice whereas long-term recognition memory is unaffected by neuronal Ldha modification at both ages

(A) Spontaneous Object Recognition (SOR) paradigm for testing recognition memory with a 24 h delay between sample and choice phase. See also Figure S14A.

(B and C) 24 h recognition memory in SOR was retained in both neuronal Ldha induction and age-matched control mice at young (B; 4.5 months; n = 18–44; phase effect: F(1, 120) = 86.24, p < 0.0001) and old (C; 11 months; n = 20–27; phase effect: F(1, 45) = 27.96, p < 0.0001) age.

(D and E) 24 h recognition memory in SOR was retained in both neuronal Ldha KO and age-matched control mice at young (D; 5.5 months; n = 18–23; phase effect: F(1, 78) = 13.32, p = 0.0005) and old (E; 11 months; n = 21–24; phase effect: F(1, 43) = 18.90, p < 0.0001) age.

(F) Spontaneous Location Recognition (SLR) paradigm for testing spatial memory with low (dSLR) or high (sSLR) pattern separation with a 4 or 6 h delay between sample and choice phase. See also Figure S14B.

(G) 6 h spatial memory in dSLR was retained at young age (5 months) in both neuronal Ldha induction and age-matched control mice (n = 20–28; phase effect: F(1, 46) = 2.62, p = 0.0009).

(H) 6 h spatial memory in sSLR at differed with sex (F(1, 44) = 4.902, p = 0.0321) in young aged (5 months) neuronal Ldha induction and age-matched control mice. Memory was retained in male Ldha induction and control mice (n = 8–15; phase effect: F(1, 21) = 15.37, p = 0.0008) but was enhanced in Ldha induction female (t(11) = 3.181, p = 0.0087) mice compared to controls (t(12) = 1.361, p = 0.1985) (n = 12–13; genotype effect: F(1, 23) = 5.829, p = 0.0241).

(I) 4 h spatial memory in dSLR was retained at old age (12 months) in both neuronal Ldha induction and age-matched control mice (n = 20–27; phase effect: F(1, 43) = 24.38, p < 0.0001).

(J) 4 h spatial memory in sSLR was lost at old age (12 months) in both neuronal Ldha induction and age-matched control mice (n = 20–27; phase effect: F(1, 45) = 1.477, p = 0.2305).

(K) 4 h spatial memory in dSLR was retained at young age (6 months) in both neuronal Ldha KO and age-matched control mice (n = 21–24; phase effect: F(1, 43) = 5.679, p = 0.0217).

(L) 4 h spatial memory in sSLR was retained at young age (6 months) in both neuronal Ldha KO and age-matched control mice (n = 21–24; phase effect: F(1, 43) = 10.58, p = 0.0022).

(M) 4 h spatial memory in dSLR was retained at old age (12.5 months) in both neuronal Ldha KO and age-matched control mice (n = 21–24; phase effect: F(1, 84) = 6.953, p = 0.01).

(N) 4 h spatial memory in sSLR at old age (12.5 months) differed with sex (F(1, 80) = 7.209, p = 0.0088) in neuronal Ldha KO and age-matched control mice. Memory in KO and control was lost in males (n = 11–12; phase effect: F(1, 42) = 0.04346, p = 0.8359) but differed in females (n = 9–12; phase × genotype effect: F(1, 38) = 4.718, p = 0.0362), with memory enhanced in KO (t(8) = 3.394, p = 0.0094) and lost in control mice (t(11) = 0.5775, p = 0.5752). Comparisons across all conditions in SOR and SLR were made using a mixed-effects model with matching by phase and fixed effects presented in each graph. For comparisons with a fixed effect of genotype or phase × genotype, further comparisons between phase for each genotype were made using paired t tests with the p value reported above each comparison. Bars in each graph represent the mean.

Neuronal Ldha transgenic mice have age- and sex-dependent changes in short-term spatial memory

Spatial memory assessed using spontaneous location recognition (SLR) depends on spontaneously elicited exploration of objects and tests pattern separation ability.⁹⁷ One advantage of the SLR test of spatial memory is the ability to systematically alter the difficulty of the task and the amount of pattern separation required. Spatial memory in young and old aged neuronal Ldha induction and Ldha KO transgenic mice was measured using two configurations of SLR with either low (dSLR) or high (sSLR) pattern separation load (Figure 6F). In young neuronal Ldha induction transgenic mice, spatial memory tested with dSLR did not differ compared to age-matched control mice (Figure 6G). However, spatial memory tested with sSLR was significantly enhanced in young neuronal Ldha induction females compared to female control mice (Figure 6H). Old neuronal Ldha induction mice compared to age-matched control mice did not differ in spatial memory tested with dSLR (Figure 6I) and sSLR (Figure 6J). Notably, both neuronal Ldha induction and age-matched control mice did not retain spatial memory tested with sSLR at old age (Figure 6J). Neuronal Ldha KO mice compared to age-matched controls did not differ in spatial memory tested at young age with dSLR (Figure 6K) and sSLR (Figure 6L) as well as when tested at old age with dSLR (Figure 6M). Interestingly, while spatial memory tested with sSLR did not differ between old neuronal Ldha KO male mice compared to age-matched control males, memory was significantly improved in old female Ldha KO mice compared to age-matched control female mice (Figure 6N).

Neuronal Ldha transgenic mice have age- and sex-dependent changes in spatial learning and long-term memory

The MWM is one of the most well-established spatial learning and memory tests for rodents. This task allows for assessment of both acquisition (learning) and retrieval of memory and produces strong enough memory traces for assessment of long-term memory and memory decay over multiple days. The MWM was used to test spatial learning in addition to 24 h and 7 days long-term spatial memory in young and old aged neuronal Ldha induction and KO mice (Figure 7A). Neuronal Ldha induction compared to age-matched control mice did not differ in spatial learning at young age (Figure 7B), but neuronal Ldha induction mice exhibited reduced spatial learning in males only at old age (Figure 7C). Neuronal Ldha KO mice compared to age-matched control mice also had no difference in spatial learning at young age (Figure 7D), but at old age female KO mice exhibited improved spatial learning compared to control mice (Figure 7E). Young neuronal Ldha induction mice had sex-dependent changes in long-term spatial memory compared to age-matched controls (Figures 7F and 7H). Female neuronal Ldha induction mice had 24 h (Figure 7FI) and 7 days (Figure 7HII) memory deficits compared to controls, whereas males had an improvement that was only detectable at 24 h (Figure 7FII). Neuronal Ldha induction compared to age-matched control mice did not differ in 24 h (Figure 7G) or 7 days (Figure 7I) long-term spatial memory at old age. Young neuronal Ldha KO mice also had sex-dependent changes in long-term spatial memory compared to age-matched controls (Figures 7J and 7L). Males had a memory deficit evident only at 7 days (Figure 7L) and not at 24 h (Figure 7J), while females did not differ from controls at either time points (Figures 7J and 7L). Neuronal Ldha KO compared to age-matched control mice did not differ in 24 h (Figure 7K) or 7 days (Figure 7M) long-term spatial memory at old age.

Figure 7.

Spatial learning and long-term memory are both differentially impacted by neuronal Ldha induction and knockout depending on age and sex

(A) Morris water maze paradigm for testing spatial learning during training days and long-term spatial memory at 24 h and 7 days probe trials.

(B) Spatial learning in young (6 months) neuronal Ldha induction was retained equally in transgenic and age-matched control mice for both sexes. n = 9–12.

(C) Spatial learning in old (12.5 months) neuronal Ldha induction compared with age-matched control mice was worse in males only (genotype effect: F(1, 19) = 5.939, p = 0.0248). n = 8–13.

(D) Spatial learning in young (7 months) neuronal Ldha KO compared with age-matched control mice was retained equally for both sexes. N = 5–7.

(E) Spatial learning in old (13 months) neuronal Ldha KO compared with age-matched control mice was improved in females only (genotype effect: F(1, 19) = 5.850, p = 0.0258). n = 9–12.

(F) 24 h long-term spatial memory in young (6 months) neuronal Ldha induction compared to age-matched control mice was decreased in females for percent time in target quadrant (I; sex × genotype effect: F(1, 38) = 7.430, p = 0.0096; t(21) = 2.125, p = 0.0457) and increased in males for mean distance from platform (II; sex × genotype effect: F(1, 38) = 7.424, p = 0.0097; t(17) = 2.594, p = 0.0189). n = 9–12.

(G) 24 h long-term spatial memory in old (12.5 months) neuronal Ldha induction compared to age-matched control was retained equally for percent time in target quadrant and mean distance from platform (I + II) for both sexes. n = 8–13.

(H) 7 days long-term spatial memory in young (6 months) neuronal Ldha induction compared with age-matched control mice was retained equally for percent time in target quadrant (I) and decreased in females only for mean distance from platform (II; sex × genotype effect: F(1, 38) = 4.459; t(21) = 2.084, p = 0.0496). n = 9–12.

(I) 7 days long-term spatial memory in old (12.5 months) neuronal Ldha induction compared with age-matched control mice was retained equally for percent time in target quadrant and mean distance from platform (I + II) for both sexes. n = 8–13.

(J) 24 h long-term spatial memory in young (7 months) neuronal Ldha KO compared to age-matched control mice is retained equally for percent time in target quadrant and mean distance from platform (I + II) for both sexes. n = 6–7.

(K) 24 h long-term spatial memory in old (13 months) was retained equally for percent time in target quadrant and mean distance from platform (I + II) for both sexes. n = 9–12.

(L) 7 days long-term spatial memory in young (7 months) neuronal Ldha KO compared to age-matched control was decreased in males only for percent time in target quadrant (I; t(11) = 3.069, p = 0.0107) and mean distance from platform (II; t(11) = 3.076, p = 0.0106). n = 6–7.

(M) 7 days long-term spatial memory in old (13 months) was retained equally for percent time in target quadrant and mean distance from platform (I + II) for both sexes. n = 9–12. For training days, comparisons were made using a mixed-effects model with Geisser-Greenhouse correction, fixed effects presented in each graph, and further comparisons between genotype for each day were made using unpaired t tests for conditions with an effect including genotype. For probe trials, comparisons were made using a two-way ANOVA, fixed effects presented in each graph, and further comparisons between genotype for each sex were made using unpaired t tests for conditions with an effect including sex for memory, learning, or locomotor ability (Figure 4). Data presented as mean ± SEM.

Neuronal Ldha transgenic mice have age- and sex-dependent changes in cognitive behavior tested in the puzzle box

The last test used to assess the cognition of neuronal Ldha induction and Ldha KO mice was the puzzle box. The puzzle box allows for screening of changes in general cognition by testing mice with consecutive trials requiring they escape from the light to the dark side of the box through increasingly difficult novel obstructions (problem solving) or through an obstruction they have passed through before (memory) (Figure 8A). This task allows for screening of deficits in medial prefrontal cortex (mPFC) changes that may alter executive function⁹⁸ and cognitive changes that may only be evident at particular levels of difficulty. No changes in habituation phase were detected in neuronal Ldha induction (Figures S15A and S15B) or neuronal Ldha KO (Figures S15C and S15D) compared with control mice. Young neuronal Ldha induction mice compared to age-matched control mice had no change in problem solving (Figure 8B) and 3 min memory (Figure 8C), but did exhibit a sex-dependent change in 24 h memory with only female transgenic mice exhibiting a deficit (Figure 8D). Conversely, old aged neuronal Ldha induction compared to age-matched control mice showed deficits in problem solving (Figure 8E) and 3 min memory (Figure 8F) but no change in 24 h memory (Figure 8G). Problem solving in young aged neuronal Ldha KO mice compared to age-matched control mice differed in the opposite direction; better for males and worse for females (Figure 8H). Young aged neuronal Ldha KO mice compared to age-matched control mice also showed a deficit in 3 min memory (Figure 8I) and a sex-dependent improvement in 24 h memory; unchanged in males and increased in females (Figure 8J). No changes in problem solving (Figure 8K), 3 min memory (Figure 8L), or 24 h memory (Figure 8M) were observed in old aged neuronal Ldha KO mice compared to age-matched controls.

Figure 8.

Cognitive behavior in puzzle box is differentially impacted by neuronal Ldha induction and knockout depending on age and sex

(A) In the puzzle box mice must pass from the light to dark side of a light-dark box with various barriers of increasing difficulty put in place during nine trials that were spread across three successive days with three trials per day. For each trial, higher latency to enter the dark side is indicative of worse performance. Habituation is exhibited during the first trial that has no barrier. Problem solving ability is exhibited during trials 2, 5, and 8 when the task difficulty is increased from the previous one on the same day. 3 min memory is exhibited during trials 3, 6, and 9 when the task is of equal difficulty to the previous one on the same day. 24 h memory is exhibited during trials 4 and 7 when the task is of equal difficulty to the previous one on the day prior. See also Figure S15.

(B) Problem solving ability was retained equally in young (7 months) neuronal Ldha induction compared to age-matched control mice. n = 13–25.

(C) 3 min memory was retained equally in young (7 months) neuronal Ldha induction compared to age-matched control mice. n = 14–24.

(D) 24 h memory in young (7 months) neuronal Ldha induction and age-matched control mice differed by sex (sex × genotype effect: F(1, 33) = 4.003, p = 0.0537; trial × sex × genotype effect: F(1, 32) = 5.143, p = 0.0302). 24 h memory in induction compared with control mice was retained equally in males, worse in females across both trial 4 and 7 (genotype effect: F(1, 18) = 10.18, p = 0.0051; trial × genotype effect: F(1, 17) = 4.713, p = 0.0444), and worse in females particularly for higher difficulty trial 7 (t(35) = 3.776, p = 0.0012). n = 5–12.

(E) Problem solving ability was worse in old (14 months) neuronal Ldha induction compared to age-matched control across trials 2, 5, and 8 (genotype effect: F(1, 27) = 4.276, p = 0.0483). n = 13–15.

(F) 3 min memory in old (14 months) neuronal Ldha induction mice was worse particularly for the intermediate difficulty trial 6 (t(79) = 2.899, p = 0.0145). n = 14–15.

(G) 24 h memory in old (14 months) neuronal Ldha induction and age-matched control differed by sex (sex effect: F(1, 25) = 5.497, p = 0.0273). 24 h memory was retained equally in induction and control mice. n = 5–9.

(H) Problem solving ability in young (7.5 months) neuronal Ldha KO and age-matched control mice differed by sex (sex effect: F(1, 41) = 11.30, p = 0.0017). Problem solving ability in KO compared with control mice was better across trials 2, 5, and 8 for males (genotype effect: F(1, 65) = 7.654, p = 0.0074) and for females (genotype effect: F(1, 19) = 4.036, p = 0.059; trial × genotype effect: F(2, 38) = 4.252, p = 0.0215). n = 9–12.

(I) 3 min memory in young (7.5 months) neuronal Ldha KO compared with age-matched control mice was better (trial × genotype effect: F(2, 86) = 4.771, p = 0.0108) particularly for the intermediate difficulty trial 6 (t(38.93) = 2.783, p = 0.0246). n = 21–24.

(J) 24 h memory in young (7.5 months) neuronal Ldha KO and age-matched control mice differed by sex (sex effect: F(1, 41) = 8.327, p = 0.0062; trial × sex: F(1, 40) = 4.371, p = 0.0429). 24 h memory in KO compared with control mice was retained equally in males and better in females (trial × genotype effect: F(1, 18) = 4.314, p = 0.0524) particularly in the higher difficulty trial 7 (t(37) = 2.455, p = 0.0375). n = 8–12.

(K) Problem solving ability in old (14 months) neuronal Ldha KO and age-matched control mice differed by sex (sex effect: F(1, 40) = 5.490, p = 0.0242). Problem solving ability was retained equally in KO and control mice. n = 9–12.

(L) 3 min memory was retained equally in old (14 months) neuronal Ldha KO compared with age-matched control mice. n = 19–24.

(M) 24 h memory in old (14 months) neuronal Ldha KO and age-matched control mice differed by sex (sex effect: F(1, 40) = 4.154, p = 0.0482). 24 h memory was retained equally in KO compared with control mice. n = 9–12. For habituation trials, comparisons made by two-way ANOVA, fixed effects presented in each graph, and unpaired t test between genotypes. For problem solving, 3 min memory and 24 h memory trials, comparisons made using a mixed-effects model, fixed effects presented in each graph, with Geisser-Greenhouse correction for problem solving and 3 min memory and Šídák’s multiple comparisons tests between genotype for each trial. Data presented as mean ± SEM.

Discussion

Neuronal lactate production can affect certain types of cognition in an age- and sex-specific manner

Here it is shown that neuronal lactate metabolism is involved in different types of cognitive processing in mice depending on sex and age (Table 1). These results provide insight into the potential role of neuronal lactate metabolism on cognition in the context of age and sex, factors mostly unexplored in previous ANLS studies. Moreover, these findings indicate that neurons are capable of generating lactate independently of astrocytes. The question arises as to the fate of lactate produced within neurons. Cytosolic lactate can either be exported via MCTs or oxidized back to pyruvate to be used as a mitochondrial fuel source. Whether lactate is converted back to pyruvate in the cytosol or within mitochondria has been a controversial issue. Almost 50 years ago it was demonstrated that both LDHA and LDHB isoforms are present in neuronal mitochondria.⁹⁹ In 1999, George Brooks demonstrated that mitochondrial localized LDH in rat liver, cardiac, and skeletal muscle can oxidize lactate to pyruvate to fuel the TCA cycle and respiration.¹⁰⁰ These observations have led to the intracellular lactate shuttle hypothesis, which posits that cytosolic lactate, formed during glycolysis, can be transported into mitochondria where it is converted to pyruvate by the mitochondrial oxidation complex (mLOC), in part composed of isomers of LDHA and LDHB.¹⁰¹ Thus, elevated neuronal Ldha expression could drive high levels of lactate production and subsequent lactate oxidation in mitochondria to fuel mitochondrial respiration and ATP synthesis to enable synaptic processes. Alternatively, cytosolic lactate can be exported and either used by adjacent cells as a fuel source or as a signaling molecule.^102,103,104 Within the brain, elevated extracellular lactate can affect nearly all cell types, including neurons, neural stem cells, neural progenitor cells, and pericytes.^{105,106,107,108,109,110} Thus, the changes in cognition resulting from either elevated or repressed neuronal Ldha expression may arise from altered lactate levels affecting metabolism or signaling in either a cell autonomous manner or through the action of secreted lactate affecting adjacent cells.

Table 1.

Summary of cognitive test results for neuronal Ldha induction and knockout mice

Test type	Measurement	Induction		Knockout
		Young	Old	Young	Old
SOR	24 h recognition memory	–	–	–	–
SLR	4/6 h spatial memory – easier – dSLR	–	–	–	–
	4/6 h spatial memory – harder – sSLR	-♂,↑♀	–	–	-♂,↑♀
MWM	Spatial learning	–	↓♂,-♀	–	-♂,↑♀
	24 h spatial memory – percent target	-♂,↓♀	–	–	–
	24 h spatial memory – mean platform distance	↑♂,-♀	–	–	–
	7 days spatial memory – percent target	–	–	↓♂,-♀	–
	7 days spatial memory – mean platform distance	-♂,↓♀	–	↓♂,-♀	–
Puzzle box	Problem solving	–	↓	↑♂,↓♀	–
	3 min memory	–	↓	↑	–
	24 h memory	-♂,↓♀	–	-♂,↑♀	–

One important finding of this study is that modifying expression of Ldha in neurons has a female-specific impact on sSLR memory that differs with age; memory was improved by neuronal Ldha induction in young mice and by neuronal Ldha KO in old mice. The higher level of pattern separation required for memory encoding in the sSLR over the dSLR test variant is a reflection of the sSLR being particularly sensitive to changes in brain-derived neurotrophic factor (BDNF) and adult hippocampal neurogenesis.^{111,112,113,114,115,116} All steps of adult hippocampal neurogenesis progression promote various forms of cognition,^{113,115,117,118,119,120,121,122,123,124,125,126,127} including pattern separation, through production of new neurons at different maturation stages.^{128,129,130,131,132} BDNF can support sSLR memory through enhanced synaptic plasticity via activation of receptors on mature neurons^133,134,135 or highly excitable immature neurons,^{136,137,138,139,140} and through promotion of adult hippocampal neurogenesis.^141,142,143

Lactate has been shown to promote Bdnf expression through neuronal silent information regulator 1 (sirtuin 1/SIRT1) signaling,^38,144 N-methyl-D-aspartate (NMDA) receptor potentiation,^17,109,145 or hydroxycarboxylic acid receptor 1 (HCAR1/GPCR81) receptor activation.^{146,147,148,149,150,151,152,153} Alternatively, lactate may trigger neural stem, precursor, or progenitor cell proliferation in the hippocampus directly through pentose phosphate pathway metabolic reprogramming¹¹⁰ or zinc-dependent anaphase promotion.¹⁵⁴ Therefore, neuronal-generated lactate may be exported to promote an increase in BDNF levels and adult hippocampal neurogenesis thereby contributing to an increase in sSLR memory in young female neuronal Ldha induction mice. Moreover, enhanced sSLR memory in young females but not males may be due to higher basal or estrogen-induced^155,156,157 expression of hippocampal BDNF¹⁵⁸ and NMDA receptors.¹⁵⁹ Further investigation is required to determine if elevated neuronal lactate production in females contributes to increased BDNF production and altered neurogenesis to facilitate pattern separation memory.

Neuronal Ldha is detrimental to cognition in aged mice

Although neuronal Ldha induction enhanced sSLR memory in young females, elevated neuronal lactate production had no effect on sSLR memory at old age in both sexes. In contrast, neuronal Ldha KO resulted in improved sSLR memory and MWM learning in old female mice. BDNF levels,^160,161,162 adult hippocampal neurogenesis,^{115,163,164,165,166,167} and cognition^115,163,168 typically decline with age. In addition, stem cell exhaustion has been identified as a hallmark of aging.^169,170 Elevated Ldha expression is associated with proliferation of various adult somatic stem cells throughout the body.^171,172,173 In vitro studies have shown that proliferative neural stem cells express high levels of glycolytic enzymes, including Ldha, and secrete elevated levels of lactate.¹⁷⁴ Furthermore, addition of lactate to the media of cultured neural progenitor cells triggers increased proliferation.¹¹⁰ As described previously, high pattern separation tasks such as sSLR memory in young female neuronal Ldha induction mice may have benefited from a lactate-dependent increase in adult hippocampal neurogenesis. In contrast, enhancement of sSLR memory at old age in female neuronal Ldha KO mice may have occurred by preventing lactate induced proliferation and exhaustion of the pool of hippocampal stem cells. Alternatively, chronically increased hippocampal lactate may impede non-neurogenesis specific cognition with age. For example, hippocampal KO of the pyruvate dehydrogenase E1 component alpha subunit (Pdha1) causes lactate accumulation, altered monocarboxylate transporter levels, and severe impairment of spatial learning and memory in the MWM, but not recognition memory.³⁷ In addition, chronic systemic chemical inhibition of glycogen breakdown, and subsequent lactate production, impairs memory in young mice^{17,21,22,28,29,52} but improves long-term potentiation in hippocampal slices¹⁷⁵ and short-term object recognition and location memory in old mice.⁵⁰ Interestingly, here it was observed that protein levels of ectopically expressed Ldha increased with age with a concomitant increase in intracellular lactate, as assessed by GC-MS. A recent study revealed that elevated brain lactate levels are linked to poorer working memory performance in multiple mouse models of neuropsychiatric disorders.¹⁷⁶ Thus, elevated neuronal lactate production with age may contribute to cognitive impairment in older mice.

Memory enhancement in old female neuronal Ldha KO mice could also be attributed to non-BDNF dependent mechanisms, such as increased LDHB-catalyzed lactate oxidization-dependent alterations in expression of immediate-early genes and synaptic plasticity genes.^109,177 There could be a shift with age toward neuronal LDHB dependent ANLS support within the hippocampus for cognitive processing. In fact, a recent finding saw global KO of Ldhb resulted in Y-maze spontaneous alternation and MWM learning and memory deficits along with apoptosis, neurodegeneration, and mitochondrial dysfunction typically associated with age-related cognitive decline.²³ Hypothetically, a metabolic shift toward ANLS supported neuronal metabolism with age may normally be enabled without changes in expression of either LDH isoform but rather the redistribution of LDHA or LDHB between neuronal processes and the soma, as neurons have been shown to localize specific metabolic enzymes to different subcellular compartments. For instance, Caenorhabditis elegans neurons under energetic stress conditions localize glycolytic enzymes toward the synapse¹⁷⁸ and hippocampal neurons in aged mice localize LDHA away from the soma.¹⁷⁹

LDHA was recently found to be enriched on axonal vesicles and is required to provide ATP for fast axonal vesicular transport.¹⁸⁰ In addition, the preferential conversion of pyruvate to lactate enabled by LDHA allowed for the concomitant recycling of the reduced form of nicotinamide adenine dinucleotide (NADH) into the oxidized form (NAD+) on vesicles.¹⁸⁰ With age, excess lactate produced by neurons, or imported from astrocytes, might block glycolytic reactions from taking place on vesicular membranes, as high lactate levels would inhibit LDHA and potentially interfere with axonal vesicular transport and synaptic function.

Numerous studies support opposing effects on the age-related dependence on ANLS and neuronal LDHB-dependent lactate oxidization. In a previous study LDHA and LDHB expression were found to decrease within the frontal cortex of mice with age.⁶⁹ Moreover, LDHA protein was detected in neurons in the brain and elevated LDHA levels correlated positively with memory in aged mice.⁶⁹ Others have found that the glycogenolysis enzyme glycogen phosphorylase increasingly localizes to hippocampal neurons with age¹⁷⁹ and KO of glycogen synthase in neurons or neural precursor cells in young mice lowers the capacity for learning in an operant conditioning test.^181,182 The discovery that glycogenolysis is not astrocyte-specific necessitates reinterpretation of previous studies suggesting that the ANLS supports cellular processing underlying cognition by showing that lactate rescues cognitive deficits caused by non-cell-type specific chemical inhibition of glycogenolysis in young animals.^{17,19,21,22,24,28,29,39,45,52} Similarly, the possibility that neuronal glycogenolysis modulates cognition makes it difficult to decipher which cell type to attribute the improvement in memory recently found in old mice treated systemically with the glycogen phosphorylase inhibitor, BAY U6751.⁵⁰ Active neuronal glycogen metabolism also suggests that glycogen may sometimes fuel synaptic functions that require glycolysis, such as presynaptic action potentials,¹⁸³ vesicle cycling,¹⁸⁴ and neurite outgrowth.⁸¹ Furthermore, chemical or genetic inhibition of glucose transport in the dorsal hippocampus reduces contextual fear conditioning learning and memory in young mice,¹⁸ while neuronal KO of a glucose transporter, Glut3, or a pyruvate-producing glycolytic enzyme, pyruvate kinase (Pkm1), causes active place avoidance learning deficits with age.⁷⁷ Of note, age-related cognitive deficits found in neuronal Pkm1 KO mice were restricted to females, much like the female-only cognitive benefits observed here in old neuronal Ldha KO mice. Hence, neuronal glycolysis or glycogenolysis, independently or in addition to the ANLS, may also be important for retention of certain cognitive abilities depending on age. The findings presented here strongly warrant further study on the connection between cognition and age-related changes in cell-type and subcellular localization of lactate metabolism and ANLS-related processes. Overall, it would appear that cognition with age becomes sensitive to changes in expression of the lactate-producing enzyme LDHA in neurons whereby increases are detrimental and decreases are beneficial.

Potential sex differences in ANLS involvement in cognition

In the present study, multiple sex differences in cognition were identified following altered neuronal Ldha expression. This observation is important because previous studies that implicate the ANLS in cognitive processes used exclusively males or failed to report whether sex differences were investigated. Numerous studies have provided experimental evidence in young vertebrates to support the involvement of ANLS in cognition, yet only five of those include females.^{37,47,51,52,53} Therefore, the extent to which sex differences affect ANLS-mediated cognition is quite unclear.

When assessing MWM long-term memory in young mice, it was observed that neuronal Ldha induction enhanced memory up to 24 h in males and decreased memory up to 7 days in females, whereas neuronal Ldha KO reduced memory in males after 7 days and did not impact female memory. Unlike other cognitive behavior paradigms employed in this study, the MWM is an inherently stressful assay. Levels of stress response hormones,¹⁸⁵ proteins such as the murine glucocorticoid corticosterone^186,187 and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) glutamate receptors, respectively, are higher in females compared to males in the basolateral amygdala, a fear- and stress-mediating brain region.¹⁸⁶ Moreover, stressors tend to have opposite effects on male and female hippocampal-dependent cognition, negative and positive, respectively.^185,187,188 Therefore, if neuronal expression of Ldha within the hippocampus mediates stress-related sex differences in memory, then this could explain the sex differences detected in MWM long-term memory. Increased neuronal Ldha may have alleviated handling and swimming-induced stress in male MWM memory while preventing the same stressors from enhancing MWM memory in females. In support of this hypothesis, decreased anxiety-like behavior, time on dark side of the light-dark box, was observed only in male neuronal Ldha induction mice. Neuronal Ldha KO may have exacerbated stress-related lowering of male MWM memory, as increased anxiety-like behavior was observed in male neuronal Ldha KO mice.

In contrast to sSLR memory, which has been closely tied to adult hippocampal neurogenesis, MWM long-term spatial memory involves hippocampal cellular processes without necessarily depending on adult hippocampal neurogenesis unless the task is altered to require increased pattern separation, cognitive flexibility, or difficulty.^{118,119,122,124,129,189,190} Dendritic spine formation is one such cellular process that occurs in the hippocampus. Although the ANLS has been implicated in learning-induced increases in hippocampal²⁸ and cortical³⁵ dendritic spine density, these studies were conducted using males only. Moreover, CA1 region dendritic large spine density¹⁵⁹ is higher in females compared to males. The overlap between lactate- and estrogen-related metabolic and signaling pathways involved in learning-dependent formation and maintenance of dendritic spines may explain why in females, compared to males, lactate has an opposite effect on some memory related neuronal processes. Both lactate and estrogen can influence dendritic spine formation through modification of mitochondrial activity and biogenesis. Lactate provides fuel for mitochondrial activity, through oxidative conversion into pyruvate by LDHB,^{73,191,192,193} and promotes mitochondrial biogenesis by increasing peroxisome proliferator activated receptor gamma coactivator α (PGC-1α) expression, through raising BDNF and cyclic adenosine monophosphate (cAMP) response element binding protein (CREB).^38,194 Estrogen can modulate mitochondrial activity by binding to intracellular estrogen receptors that, in turn, bind estrogen response elements within the promoter regions of mitochondrial genes.^{195,196,197,198,199,200,201,202} Estrogen also promotes mitochondrial biogenesis through nuclear estrogen- and estrogen-related receptors that induce expression of Bdnf^156,157,203 and PGC-1α.^204,205,206 Moreover, estrogen-related receptors colocalize with estrogen receptors in the hippocampus,²⁰⁷ and have been found to upregulate Ldhb and neuronal mitochondrial oxidative metabolism required for long-term potentiation and cognition.²⁰⁸ Therefore, hippocampal dendritic spine regulation required for MWM long-term memory in young female mice may depend on estrogen-mediated neuronal mitochondrial functions that are facilitated by the ANLS. In contrast to females, hippocampal neurons in young male mice may have the metabolic flexibility to utilize neuronal generated lactate to promote dendritic spine formation and improve MWM long-term memory. Future studies are required to determine whether altered neuronal Ldha expression affects MWM memory in young mice differently between sexes due to overlapping roles of estrogen and the ANLS in modulating synaptic plasticity via mitochondria or opposing effects of stress on the hippocampus in males and females.

Metabolic alterations arising from altered neuronal Ldha expression

GC-MS analysis of hippocampal tissue extracts revealed several metabolites that varied between experimental conditions and sex. Ascorbic acid was elevated in old male neuronal Ldha induction mice and young female Ldha induction mice relative to age-matched control mice. Intracellular ascorbic acid inhibits glucose consumption and stimulates lactate uptake in neurons.²⁰⁹ In addition, glutamate (glutamic acid) stimulates ascorbic acid release from astrocytes, which is subsequently transported into neurons. Both glutamic and aspartic acid decreased with age in control male mice, and to a lesser extent female control mice. However, the decline in glutamic and aspartic acid with age was less pronounced in male Ldha KO mice. Aspartic and glutamic acid are also important metabolites in the brain that are involved in the malate-aspartate shuttle (MAS) connecting glycolysis to the TCA cycle.²¹⁰ Due to the impermeability of the inner mitochondrial membrane to NAD+ and NADH, NADH generated during glycolysis depends on the indirect transfer of reducing equivalents into the mitochondria via the MAS. Lactate transported into the mitochondrial inner membrane space may be converted to pyruvate by mitochondrial localized LDHA with the concurrent reduction of NAD+ to NADH. The regeneration of NAD+ within the inner membrane space to maintain LDHA activity is achieved by the coordinated action of the MAS and the glutamate/aspartate carrier.²¹¹ Thus, it is possible that perturbations to neuronal lactate metabolism promote compensatory changes in other metabolites to maintain mitochondrial activity required for synaptic processes. In keeping with the sex differences observed in other measures, the top candidates identified by GC-MS often seem to differ greatly by sex. Further studies are required to determine how these metabolites connect with lactate metabolism and contribute to age-related cognitive decline in both sexes. Overall, our results suggest that sex is an important mitigating factor in neuronal lactate metabolism involved in spatial memory.

Various types of cognition differ in their dependence on neuronal lactate metabolism

In this study, SOR long-term memory was assessed using a Y-maze apparatus, instead of an open field, to restrict mice from using spatial cues for object discrimination, thereby lowering the spatial complexity of the task, reducing requirement for hippocampal spatial processing, and increasing the specificity for need of protein synthesis dependent activity in the perirhinal cortex.^{94,96,212,213,214,215} Altered expression of neuronal Ldha had no impact on SOR memory regardless of age and sex. These results suggest a limited role for neuronal lactate metabolism in the perirhinal cortex. Moreover, to the extent that hippocampal function still plays a role in SOR memory when tested in a Y-maze, the results presented here are in line with a recent study that showed hippocampal injection of oxamate, a potent LDH inhibitor, had no effect on novel object recognition memory tested in an open field.³¹ In contrast to SOR tested in a Y-maze, the puzzle box cognitive task necessitates mice navigate through a larger arena with more surrounding visual input toward a barrier that requires various levels of physical exertion to overcome. The complexity of this task makes the puzzle box a crude measure of mouse executive function. Puzzle box 3 min memory and problem solving are not analogous to behavioral outcomes in the other cognitive tasks. However, the female-specific long-term memory deficit of young neuronal Ldha induction mice in the puzzle box mirrors the findings in the MWM. The results from the puzzle box suggest neuronal lactate metabolism may be involved in executive function, but deeper examination using tasks designed specifically for testing cognitive flexibility and more strictly evaluate problem solving ability are required to make more robust conclusions about the ANLS involvement in this type of cognition.

The results presented in this study demonstrate that genetic alteration of neuronal Ldha expression in mice promotes cognitive changes that differed by age and sex. Using a variant of spatial memory testing requiring high pattern separation, female-specific cognitive enhancements were identified in young mice with raised neuronal Ldha expression and in old mice with ablated neuronal Ldha expression. Across multiple measures of cognition at old age, induced neuronal Ldha expression caused deficits whereas neuronal Ldha KO caused improvements. These findings provide evidence supporting a potential age- and sex-related shift in the ability of neurons to utilize lactate for the purpose of supporting cognition. This study bolsters past findings showing the importance of lactate metabolism for enabling cognitive processes and highlights the need for future investigation to include females and older animals.

Limitations of the study

The genetic manipulations made in this study (Figure 9) were restricted to CaMKIIα expressing neurons, which represent a broad category of neurons throughout the brain. There may be non-CaMKIIα expressing neurons that were unaffected by the conditional induction and KO of Ldha yet rely on cognition-relevant lactate metabolism. Moreover, genetic manipulations were not restricted to any single brain region and occurred chronically after Ldha induction or ablation was induced. Acute manipulation of neuronal lactate metabolism in different brain regions using viral vectors may allow for less compensation and yield different behavioral outcomes. Lastly, there are several forms of post-translational modifications, including phosphorylation,^{216,217,218,219} acetylation,^220,221,222 succinylation,²²³ methylation,^224,225 and lactylation^226,227 that can affect the activity of LDHA but were not assessed in this study. Moreover, other factors that can also affect LDHA activity include: post-transcriptional regulation of messenger ribonucleic acid (mRNA) stability^228,229,230; direct binding to long non-coding RNA (lncRNA)^231,232; and protein-metabolite interaction with various keto acids related to pyruvate and lactate, nucleotides and dinucleotides, adenosine nucleotide derivatives, and coenzyme A (CoA) derivatives.²³³

Figure 9.

Neuronal Ldha induction and KO mouse experimental timelines

Experimental timeline for neuronal Ldha induction mice (A) and KO mice (B). Branches in each timeline indicate utilization of separate cohorts with different trajectories.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Experimental models: organisms/strains
Mouse: TRE-LDHA	This paper	RRID:MGI:7645158
Mouse: CaMKII-tTA	The Jackson Laboratory	RRID:IMSR_JAX:007004
Mouse: C57BL/6NCrl	Charles River	RRID:IMSR_CRL:027
Mouse: Ldhaflox	The Jackson Laboratory	RRID:IMSR_JAX:030112
Mouse: CaMK2α-CreERT2	The Jackson Laboratory	RRID:IMSR_JAX:012362
Antibodies
Mouse anti-HA.11 epitope tag	BioLegend	BioLegend Cat# 901513, RRID:AB_2565335
Rabbit anti-LDHA	Cell Signaling Technology	Cat# 2012, RRID:AB_2137173
Rabbit anti-LDHB	Proteintech	Cat# 19988-1-AP, RRID:AB_10638780
Goat anti-mouse, HRP conjugated	Millipore Sigma	Cat# AP130P, RRID:AB_91266
Goat anti-rabbit, HRP conjugated	Millipore Sigma	Cat# AP132P, RRID:AB_90264
Goat anti-mouse, affinity-purified fab fragment	Jackson ImmunoResearch	Cat# 115-007-003, RRID:AB_2338476
Rabbit anti-MAP2	Abcam	Cat# ab32454, RRID:AB_776174
Rabbit anti-LDHA	Proteintech	Cat# 19987-1-AP, RRID:AB_10646429
Mouse anti-CaMKIIα	Cell Signaling Technology	Cat# 50049, RRID:AB_2721906
Goat anti-rabbit, Alexa Fluor 647	ThermoFisher Scientific	Cat# A-21244, RRID:AB_2535812
Goat anti-mouse, Alexa Fluor 568	ThermoFisher Scientific	Cat# A-11031, RRID:AB_144696
Goat anti-rabbit, Alexa Fluor Plus 800	ThermoFisher Scientific	Cat# A32735, RRID:AB_2633284
Software and algorithms
GraphPad Prism	GraphPad Software	https://www.graphpad.com, RRID:SCR_002798
NIS-Elements AR	Nikon Instruments	https://www.nikoninstruments.com/Products/Software, RRID:SCR_014329
Image Lab	Bio-Rad	http://www.bio-rad.com/en-us/sku/1709690-image-lab-software, RRID:SCR_014210
ANY-maze	Stoelting	https://www.any-maze.com, RRID:SCR_014289
LECO ChromaTOF-Sync software	LECO	https://www.leco.com/product/chromatof-sync, RRID:SCR_023077
Metaboanalyst	Jeff Xia Lab, McGill University	https://www.metaboanalyst.ca/, RRID:SCR_015539
Deposited data
Raw and analyzed data	This paper	https://doi.org/10.5683/SP3/PWONIY

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Robert C. Cumming (rcummin5@uwo.ca).

Materials availability

This study did not generate new unique reagents. The TRE-LDHA mouse line is available upon request from the lead contact.

Data and code availability

• •All raw data and output from statistical analyses, including metabolomic data, have been deposited at the Borealis data repository and are available as of the date of this publication. Accession numbers are listed in the key resources table.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Mice

This study was performed in accordance with Canadian Council of Animal Care guidelines. All animals were handled according to protocols approved by the animal care committee (ACC) of Western University (Protocol 2011-079 and 2020-112). Male and female mice were housed individually or grouped by sex in shoebox cages in a temperature-controlled room (22-25°C) with a 12 hour light-dark cycle in a plastic house (ACRE011, Techniplast) and with ad libitum access to food (Teklad 2018, Envigo) and water. Neuronal Ldha induction transgenic mice were generated using the Tet-Off system^82,83 as described in Figure 1A. Neuronal Ldha induction transgenic and littermate control mice were bred by crossing mice containing a transgene with a neuronal calcium-calmodulin-dependent kinase II (CaMKIIα) promoter driving tTA expression (CaMKII-tTA; 007004, The Jackson Laboratory)²³⁴ and mice containing a transgene with a TRE promoter driving HA-tagged Ldha expression (TRE-LDHA). The mouse Ldha cDNA sequence fused to an HA tag sequence on the 3’ terminal was ligated into the multiple cloning site of the pTRE-Tight Vector (631059, Clontech) and linearized using the restriction enzyme Xho1. Pronuclear injection of the linearized insert was conducted by London Regional Transgenic and Gene Targeting Facility, Victoria Research Laboratories, London, ON to generate the TRE-LDHA line. CaMKII-tTA and TRE-LDHA mice were maintained hemizygous on a C57BL/6NCrl background (027, Charles River). Offspring from crossing CaMKII-tTA and TRE-LDHA mice were genotyped using endpoint polymerase chain reaction (PCR). Offspring carrying both transgenes were designated neuronal Ldha induction transgenic mice whereas those carrying only one transgene or neither were designated control mice. Neuronal Ldha induction and littermate control mice were provided 200 mg/kg doxycycline in their diet (S3888, Bio-Serv) during embryonic and early age development to prevent tTA-induced HA-tagged Ldha expression in transgenic until 1.5 months of age.²³⁵ Neuronal Ldha knockout (KO) mice were generated using the ligand-activated site-specific Cre-lox system^86,87 as described in Figure 2A. Neuronal Ldha KO mice and littermate control mice were bred by crossing mice homozygous for exon 3 loxp-flanked Ldha (Ldha^flox; 030112, The Jackson Laboratory)¹⁷³ and hemizygous for a transgene with a neuronal CaMKIIα promoter driving CreERT2 expression (CaMK2α-CreERT2; 012362, The Jackson Laboratory)²³⁶ to mice homozygous for Ldha^flox without the CaMK2α-CreERT2 transgene. The CaMK2a-CreERT2;Ldha^flox and Ldha^flox mice were maintained on the mixed C57BL/6N;C57BL/6J genetic background which the Ldha^flox were provided on from The Jackson Laboratory. Offspring from crossing CaMK2α-CreERT2;Ldha^flox to Ldha^flox mice were genotyped using endpoint PCR. Offspring carrying the CaMK2α-CreERT2 transgene were designated neuronal Ldha KO mice whereas those without it were designated control mice. Neuronal Ldha KO and littermate control mice were provided 500 mg/kg tamoxifen in their diet (TD130857, Envigo) at 4-5 months of age for a month to activate CreERT2 recombination-induced Ldha knockout²³⁷ and a similar tasting diet without tamoxifen (TD0757, Envigo) for five days prior. Transgenic and age-matched controls underwent behavioural testing simultaneously at young and old ages at least one month after having the transgenic system induced (Figure 9). Cognitive behaviour in spontaneously elicited object-based tasks, and the puzzle box were repeatedly tested at young and old age. Cognitive behaviour in the Morris Water Maze was tested in separate cohorts for young (6-7 months) and old age (13 months). Locomotor ability on the rotarod and anxiety-like behaviour in the light-dark box were tested only at old and young age, respectively. In vivo ¹H-MRS analysis was performed on mice only after completing behavioural testing. Mice were euthanized with carbon dioxide, cardiac perfused with phosphate buffered saline (PBS) containing phenylmethylsulfonyl fluoride (PMSF) protease inhibitor (P7626, Millipore Sigma) and sodium orthovanadate phosphatase inhibitor (S6508, Millipore Sigma), and then had brains harvested for further analysis.

Method details

Western blotting

Brain tissues dissected from particular brain regions in the right hemisphere of each mouse were extracted using lysis buffer containing 250mM sucrose (S0389, Millipore Sigma), 50mM tris (BP152, ThermoFisher Scientific), 25mM potassium chloride (P4504, Millipore Sigma), 1% triton X (T9284, Millipore Sigma), 0.5mM PMSF (P7626, Millipore Sigma), 1X Halt™ protease inhibitor cocktail (87786, ThermoFisher Scientific), and 0.1mM sodium orthovanadate (S6508, Millipore Sigma). Brain tissue was dounce homogenized, centrifuged, and protein concentration in the supernatant was determined using a detergent compatible assay (5000111, Bio-Rad). Protein extracts were combined with a bromophenol blue based loading buffer and boiled for five minutes. For each experiment, an equal fraction of samples from each replicate were combined to create a pooled sample. Protein extracts and the pooled sample were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a polyvinylidene fluoride (PVDF) membrane blot. For each gel, protein standards (1610373, Bio-Rad) were run alongside the protein samples. Total protein level for each blot was quantified from images generated using staining with 0.1% Ponceau S (P3504, Millipore Sigma) in 5% glacial acetic acid. Blots were blocked with 1% w/v bovine serum albumin (BSA) in tris buffered saline containing 0.1% tween 20 (TBST) and 3% w/v milk at room temperature, probed with primary antibodies in at 4°C overnight, and secondary antibodies at room temperature for 2 hours. Primary antibodies were made in TBST with 0.01% sodium azide, and blots were washed in TBST before and after each probe. Primary antibodies included mouse anti-HA.11 epitope tag (901513, BioLegend; 1:1000), rabbit anti-LDHA (2012S, Cell Signaling; 1:1000), and rabbit anti-LDHB (19988-1-AP, Proteintech; 1:3000). Secondary antibodies conjugated with horseradish peroxidase (HRP) included goat anti-mouse (AP130P, Millipore Sigma; 1:10000), and goat anti-rabbit (AP132P, Millipore Sigma; 1:6666). Immobilon Classico or Forte Western HRP substrate (WBLUC, WBLUF, Millipore Sigma) were used to produce chemiluminescent signals on probed blots. A ChemiDoc XRS imaging system (170-8070, Bio-Rad) was used to image chemiluminescence and Ponceaus S signal density and quantified using Image Lab software (RRID:SCR_014210, Bio-Rad). For quantification purposes, western blot band intensity was standardized to total Ponceaus S signal density for each lane then normalized to the pooled sample on each blot. Males and females were analyzed and graphed separately if there were sex-specific statistical effects.

Immunofluorescence microscopy

The left hemisphere of brains harvested from each mouse were post-fixed in 4% paraformaldehyde (15713, Electron Microscopy Sciences) in phosphate buffer for three days then stored in 70% ethanol. Fixed brains were embedded in paraffin wax and sectioned with a rotary microtome (RM2055, Leica Biosystems) at a thickness of 5μm along the sagittal plane until the polymorph layer of the hippocampal dentate gyrus was visible. Sections were baked onto glass slides (12-550-15, ThermoFisher Scientific), deparaffinized using xylene, rehydrated with successive solutions of increasingly diluted ethanol, and then subjected to heat-induced epitope retrieval with 10mM sodium citrate buffer pH 6.0 (C7129, Millipore Sigma) at 90°C for 15min. Sections were exposed to ultraviolet (UV) light to quench general autofluorescence, 1mg/ml sodium borohydride (452882, Millipore Sigma) in PBS to quench fixative-derived autofluorescence, and TrueBlack or TrueBlack Plus (23007, 23014, Biotium; 1:20, 1:50) to quench lipofuscin-derived autofluorescence. Sections were treated with goat anti-mouse affinity-purified fab fragment antibodies (115-007-003, Jackson ImmunoResearch Inc.; 1:40) and Background Sniper (BS966, Biocare Medical) for 7 minutes or Ultra Cruz (sc-516214, Santa Cruz) for 1 hour to block non-specific signal. Sections in a humidity chamber were hybridized with primary antibody at 4°C overnight, secondary antibody for 45-60 minutes at room temperature, and counter stained with 1:300 4’,6-diamidino-2-phenylindole (DAPI; D1306, ThermoFisher Scientific) in PBS for 2 minutes at room temperature. Sections were mounted under coverslips using Prolong Gold Antifade reagent (P36930, ThermoFisher Scientific). Antibodies were prepared in Dako antibody diluent (S0809, Agilent Technologies). Primary antibodies included mouse anti-HA.11 epitope tag (901513, BioLegend; 1:500), rabbit anti-MAP2 (ab32454, Abcam; 1:700), rabbit anti-LDHA (19987-1-AP, Proteintech; 1:250), mouse anti-CaMKIIα (50049, Cell Signaling; 1:250). Secondary antibodies conjugated to fluorophores included goat anti-rabbit Alexa Fluor 647 (A21244, ThermoFisher Scientific; 1:500), goat anti-mouse Alexa Fluor 568 (A11031, ThermoFisher Scientific; 1:500 / 1:400), goat anti-rabbit Alexa Fluor Plus 800 (A32735, ThermoFisher scientific; 1:350). Sections were washed with PBS between each step of the staining process from quenching onwards. Fluorescence microscopy images were taken of each section using a Nikon ECLIPSE Ti2-E microscope system with plan apochromatic lambda dry 20x objective (MRD00205, Nikon Instruments) and a monochrome backside illuminated scientific complementary metal–oxide–semiconductor image sensor (pco.edge 4.2 bi, Excelitas Technologies). The excitation and emission wavelengths (nm) used for detection of fluorescence from Alexa Fluor 568 were 554 and 595, from Alexa Fluor 647 and Alexa Fluor Plus 647 were 635 and 681, for Alexa Fluor Plus 800 were 735 and 809, for DAPI were 378 and 432, respectively. 3x3 tiled and stitched images were created for each brain region: hippocampus, cerebellum, and frontal cortex. Within NIS-Elements AR (RRID:SCR_014329, Nikon Instruments) the Clarify.ai module was used to automatically remove blur from images. Signal quantification was performed using a custom-made GA3 analysis pipeline. Thresholds based on fluorescent signal intensity were determined for each channel before being systematically applied to quantify total signal area across all images. Percent MAP2+ neurons with HA equals the area of HA and MAP2 signal overlap divided by total MAP2 signal area. Percent CaMKIIα+ neurons with LDHA equals the area of LDHA and CaMKIIα signal overlap divided by total CaMKIIα signal area. Males and females were analyzed and graphed separately if there were statistical effects including sex.

¹H magnetic resonance spectroscopy

Magnetic resonance imaging (MRI) experiments were performed on a 9.4-T/31-cm small animal MRI scanner (Agilent, Palo Alto, CA, USA) interfaced to a Bruker Avance III HD console (Bruker BioSpin Corp, Billerica, MA) and equipped with a 6-cm gradient coil of 1000mT/m strength, with software package of Paravision-6 at the Centre for Functional and Metabolic Mapping located within the Robarts Research Institute at the University of Western Ontario. A Varian 30-mm millipede volume RF coil was used for data collection. Mice were anesthetized using isoflurane maintained between 1.5% and 2.5% with an oxygen flow rate of 1−1.5 L/min through a custom-built nose cone. Animal temperature was monitored with a rectal temperature probe and respiration was monitored with a pneumatic pillow connected to a pressure transducer that was placed on the thoracic region. Body temperature was maintained at 36.9−37.1°C throughout imaging by blowing warm air over the animal using a model 1025 small-animal monitoring and gating system (SA Instruments Inc., Stony Brook, NY, USA). At the beginning of each scan, the T2-weighted anatomical images were acquired to plan the MRS voxel. This T2-weighted anatomical image was acquired using a TurboRARE2D pulse sequence (16 averages, 31 slices with slice thickness of 0.5 mm, FOV = 19.2 × 19.2 mm², matrix size = 128 × 128, in-plane resolution = 0.15 × 0.15 mm², TE = 40 ms, TR = 5.0 s, echo spacing = 10 ms, and Rare factor = 8). Subsequently, a 2 x 6 x 3 mm³ voxel was positioned to both hippocampi for MRS data acquisition using a semi-LASER (Semi-Localization by Adiabatic Selective Refocusing, provides localized spectra from rectangular voxels selected with a sequence of an 90° excitation pulse and two pairs of 180° adiabatic refocusing pulses. Four spin echoes are produced from which the last one is acquired) pulse sequence with parameters: TR/TE = 5000/136ms. Lactate was identified as an inverted doublet peak at 1.33ppm. A water spectrum with 8-average was acquired first, then a 128-averaged metabolite spectrum with water suppression was acquired. Metabolite spectra were fitted using fitMAN software²³⁸ incorporated into a graphical user interface written in the IDL (Interactive Data Language) programming language²³⁹ to determine the amplitude of lactate and creatine resonances. The lactate over creatine ratio was calculated by dividing the sum of each amplitude resonance after T2 correction.²⁴⁰ Males and females were analyzed and graphed separately if there were statistical effects including sex.

Gas chromatography-mass spectrometry

Primary metabolites were extracted with methanol-methyltertbutyl ether- water, as modified from Ding et al., (2021).²⁴¹ In brief, frozen hippocampus tissue was ground with a pestle, of which 5 mg was suspended in 225 μL of cold (-20°C) methanol (MeOH) . Cold (-20°C) methyl tert-butyl ether (MTBE; 750 μL) was added and the samples were vortexed, placed on a rotating mixer at 4°C for 30 mins, and transferred to a sonicating bath at 4°C for 15 mins. To induce phase separation, 188 μL room temperature ultra-pure water containing 0.044 mg/mL adonitol (02240, BioChemika) as an internal standard was added, and the samples were vortexed and centrifuged for 10 mins at 13,000 rpm to pellet insoluble material. For recovery of polar metabolites, two 125 μL aliquots of the MeOH-water (lower) phase were transferred to separate, clean microcentrifuge tubes.²⁴¹ After the addition of 10 μL of 0.1mg/mL mixture of n-alkane retention index standards (C12, C15, C19, C22, C26, and C30), the samples were dried under a stream of N₂. Samples were stored at -20°C until further analysis. Dried polar extracts are derivatized using a two-step process involving methoximation followed by trimethylsilylation, using a protocol adapted from Fiehn (2016).²⁴² Briefly, 20 μL of 20 mg/mL methoxyamine-HCL (226904, Millipore Sigma) in pyridine was added to the dried extract residues and the samples incubated at 30°C for 90 min. Next, 80 μL of N-methyl-N-(trimethylsilyl)triflouroacetamide (MSTFA) (69479, Millipore Sigma) was added to each sample and incubated for 30 min at 37°C. Derivatized samples were cooled to room temperature, centrifuged, and 10 μL was transferred to chromatography vials fitted with glass micro-volume inserts and diluted with chloroform to a final volume of 100 μL. Samples were immediately analyzed via GCMS. A composite sample, created by pooling equal amounts from all samples, was prepared in the same way as detailed above; the composite sample was run every 10 injections. Samples were run in two cohorts to reflect the experimental design. Each cohort included an injection of chloroform, 6-point quality control mix (Fiehn 2016),²⁴² retention index mix, and reagent blank at the start of each cohort. For calibration, a dilution series was prepared with sodium lactate (200 μg/mL – 0.0002 μg/mL) and derivatized as above. Samples (1 μL) were injected into an Agilent 7890A GC system configured with a RESTEK Rxi-5ms Low bleed GC column (30 m, 250 μm internal diameter and 0.25 μm film thickness) (709-809-508, Restek) and eluted with the following oven temperature program: initial temperature at 50°C held for 0.5 mins followed by a temperature ramp at 20°C min^-1 to a final temperature of 325°C and held at 325°C for 5.75 mins. Data acquisition was turned on after a 3-minute delay to allow the solvent to clear the system. Injector and transfer line temperatures were set to 250°C. High purity helium was used as carrier gas at a flow rate of 1 mL min^-1. Mass data were collected over an 85-500 m/z range using a LECO Pegasus BT MS. Data produced (i.e. sample files, five reagent blanks, and five 6-point quality control mixes) were baseline corrected, deconvoluted, aligned, and peak areas were integrated using LECO ChromaTOF-Sync software (RRID:SCR_023077). To control for efficiency of derivatization, a comparative check on the 6pt – quality control mix was performed with each batch as detailed in Fiehn (2016).²⁴² Integrated peak areas were normalized between chromatograms based on peak area of adonitol. Only components with peak areas greater than the average peak area of five reagent blanks were retained. Analyte annotation of mass spectra features was based on the National Institute of Standards and Technology (NIST) library. Metabolites that could not be identified in the NIST database were numbered based on their mean retention time.

Rotarod

A rotarod (EZ-Rod, AccuScan Instruments) was used to test locomotor ability²⁴³ according to a modified protocol^244,245 depicted in Figure 4A. Mice were habituated to the testing room for at least one hour prior to beginning the protocol. On day 1, mice were placed on the rotarod for two minutes with the 3 cm diameter rod spinning at a constant 10.5 revolutions per minute and the mouse returned to the rod each time after a fall. On day 2, mice were subjected to six trials separated by 10 minute intertrial intervals with each trial involving mice being placed on the rod spinning at 4.3 revolutions per minute with the speed increased by 4 revolutions per minute every 30 seconds for a maximum of five minutes or until the mice fell off. The apparatus was cleaned with 70% ethanol between mice and between trials. A lower number of falls on day 1 was considered a measure of the increased ability of each mouse to become familiarized to the rod. Higher latency to fall was considered a measure of locomotor learning on day 2 during trial one to five and a measure of maximum locomotor ability during trial six.²⁴⁴ Data for males and females were graphed separately for day 1 and 2 if there was an effect including sex on either day.

Light-dark box

The light-dark box test^246,247 was used to test anxiety-like behaviour using a modified protocol.^248,249 The light-dark box was built in-house using white acrylic on the light side and black acrylic on the dark side. The dimensions of the box were 27.5 cm high walls all around, 27.4 x 59.5 cm light side, and 27.4 x 15.2 cm dark side covered from light with black acrylic. An open hole in the barrier between the light and dark side was 3 x 4.2 cm. Corncob bedding (Teklad 7097, Envigo) and a plastic shelter were placed in the dark side. Mice were placed in the light side at the end farthest from the dark side with the light side brightly lit. Mice were provided five minutes to freely roam the apparatus while video recorded (C920, Logitech) and analyzed manually. The light side was cleaned with ethanol between mice. Anxiety-like behaviour can be difficult to dissociate from impulsivity.^{250,251,252,253} Latency to enter the dark side the first time was considered a combination of impulsivity and anxiety-like behaviour whereas total time spent on dark side was considered a more reliable measure of anxiety-like behavior.²⁵⁴ Mice that never entered the dark side were excluded. Data from both measures of latency to enter the dark side and time in dark side were transformed logarithmically.

Spontaneously elicited object-based cognitive behaviours

The two spontaneously elicited objected-based cognitive behaviours used in this study include spontaneous object recognition (SOR) to test 24 hour long-term recognition memory based on a protocol adapted from others^96,255 and spontaneous location recognition (SLR) to test 4 or 6 hour short-term spatial memory based on a protocol adapted from others.^97,111,112 Mice were subjected to SOR prior to SLR at young and old age. Each mouse was familiarized to being handled by the experimenter for one minute per day for at least 14 days prior to testing. To encourage increased spontaneous exploratory behaviour, mice were food deprived down to 85-90% body weight for one week prior to testing and maintained at that weight during the SOR protocol. Moreover, mice were tested during the dark phase of their light cycle for both SOR and SLR also to increase exploratory behaviour. The SOR protocol was conducted using a white acrylic Y-shaped arena, 32.5 cm high walls and 6.5 cm width arms, under red light with one trial per mouse per day for five days. Mice were habituated to the empty SOR arena for 10 minutes each on day 1 to 3 and provided objects at the end of two arms of the Y during the 10 minute sample phase on day 4 and 5 minute choice phase on day 5, as depicted in Figure 6A. Two identical objects were used in the sample phase. In the choice phase, the familiar object was identical to that used in the sample phase and the novel object differed in shape and colour. Two separate sets of objects were used at each age tested (Figure S14A). The object type used in the sample phase and the left-right position of the novel object was counterbalanced within each group. 70% ethanol was used to clean objects between each mouse and clean the arena after any mouse had urinated or defecated. The SLR protocol was conducted over seven days using a circular arena, 33 cm high walls and 39 cm diameter, with corncob bedding (Teklad 7097, Envigo) coating the floor, different white spatial cues placed at a height of 48 cm on three of the four 122 cm high black walls positioned 5 cm outside the arena. Mice were habituated to the empty SLR arena for 10 minutes per day on day 1 to 5 and tested with objects in the arena during a 10 minute sample phase and 5 minute choice phase separated by a 4 or 6 hour delay on day 6 and 7 as depicted in Figure 6F. On the first day of habituation, the empty SLR arena functions as an open field in which thigmotaxis anxiety-like behaviour was measured as the time spent near the wall in the outer zone of the arena (Figure 5A) in a similar manner used by others in the past.^116,256 In the sample phases, three identical objects were placed in a triangular formation with their locations relative to the spatial cues and the distance between two of the three objects varied on day 6 and 7 to be either farther apart (dissimilar, dSLR) or closer together (similar, sSLR). In the choice phases, two objects, identical to those used in the sample phase, were placed such that one was in a novel location equidistant from the two object locations with variable distances apart in the sample phase prior and one in a familiar location that was identical to the third object in the sample phase prior. SLR was tested at two levels of difficulty. A higher requirement for pattern separation to distinguish between the novel and familiar location in the choice phase of sSLR compared to dSLR makes sSLR a more difficult task for mice than dSLR. Object type differed in dSLR and sSLR for each group of mice with the object type and order of testing for dSLR and sSLR counterbalanced. Different object types and locations relative to the spatial cues were used for testing at young and old age (Figure S14B). The corncob bedding was partially replenished, defecation pellets removed, and objects cleaned with 70% ethanol between each mouse. In SOR and SLR, mice were video recorded (C920, Logitech) and exploration of objects was manually scored with scorers blinded to the age, genotype, and sex of the mice. Objection exploration times were used to calculate a discrimination ratio for each mouse in the sample phase for SOR and SLR. To control for innate bias between locations in SLR and SOR, discrimination ratios were compared between sample and choice phases. SOR sample phase discrimination ratios were calculated by subtracting exploration time of the object on the side that will be the location of the familiar object in the choice phase for that mouse from exploration time of the object on the other side, with exploration time on each side normalized to total exploration time for that trial. SLR sample phase discrimination ratios were calculated by subtracting exploration time of the object in the location that will be the familiar location in the choice phase for that mouse on that day from the average exploration time of the two objects in locations that will be absent in the choice phase for that mouse on that day, with exploration time for each object normalized to total exploration time for that trial. SOR and SLR choice phase discrimination ratios were calculated by subtracting exploration of the novel object (SOR) or location (SLR) from exploration time of the familiar object (SOR) or location (SLR), with exploration time for each object normalized to total exploration time for that trial. The magnitude of increase in discrimination ratio for choice phase compared with sample phase was considered a measure of memory in both SOR and SLR. Data for males and females were analyzed and graphed separately if there were sex-specific statistical differences.

Morris Water Maze

The Morris Water Maze²⁵⁷ was used to test spatial learning and long-term memory, as well as locomotor ability, as adapted from a previous protocol²⁵⁸ depicted in Figure 7A. The testing area was arranged with a 120 cm diameter tank (San Diego Instruments) placed in the centre, different spatial cues on each of the four walls, and an adjacent holding area, out of site from inside the testing area. The tank was filled with water made opaque with white paint (SAR173696, Sargent Art Inc) and maintained at approximately 26°C for each of the 15 days in the protocol. A 15 cm diameter circular platform was placed approximately 1 cm below the surface of the water. On day one, all mice were habituated to the holding area for 1 hour remaining in their home cages. On day two, each mouse was placed on the platform in the centre of the tank and remained there for a continuous 15 seconds to become familiarized to the testing area, tank, and platform. On day three, each mouse was familiarized to swimming and escaping on an invisible platform by being placed in the water at edge of the tank and given 90 seconds to swim to the platform in the centre and remain there for a continuous 15 seconds. Days 4-7 were considered days 1-4 of training mice to find the platform in the correct target location. On every training day each mouse underwent four trials with 10 minute intertrial intervals. During intertrial intervals, each mouse was kept alone in a holding cage on a heating pad. Training day trials each involved placing a mouse in the water at the edge of the tank with the starting location corresponding to a rotating cardinal direction each trial and each day. Once in the water for a training day trial, mice were given 90 seconds to swim to the platform in the centre of an unchanging target quadrant and remain there for a continuous 15 seconds. If a mouse jumped off the platform on day one or had not found the platform in 90 seconds on day two or any of the training day trials, then at that time the mouse was directed to the platform to ensure it remained there for a continuous 15 seconds. On day eight and day fifteen, each mouse underwent the 24 hour and 7 day probe trial, respectively. Probe trials involved placing each mouse in the water at the edge of the tank and letting them swim without the platform for 60 seconds. After the 7 day probe trial, each mouse underwent a flag trial whereby the mouse was placed in the water at the edge of the tank and given 90s to find the platform in the centre of the quadrant opposite the target quadrant with a flag on it to make it visible above the water. All data from mice that failed to find the platform during the flag trial were excluded. For every trial, mice that were not in the tank during that trial and the experimenter remained in the holding area. Mouse behaviour in the tank was video recorded (C920, Logitech) and tracked using automated software (RRID:SCR_014289, ANY-Maze, Stoelting). Lower average latency and distance travelled to find the platform were measures taken on each training day interpreted as learning. Higher percentage of time spent in the target quadrant and lower mean distance from the centre of the platform area in the target quadrant were measures taken on each probe trial interpreted as memory. Higher average swimming speed on training days and swimming speed and distance travelled in probe trials were measures interpreted as locomotor ability. Data for male and female mice were analyzed and graphed separately for training days if there was a sex difference identified in learning or locomotor ability during training and for probe trials if there was a sex difference identified in memory, learning, or locomotor ability during probe. Locomotor ability was measured as average swimming speed during training days and swimming speed and distance travelled during probe trials.

Puzzle box

The puzzle box test⁹⁸ was used to test various cognitive abilities, including problem solving ability, 3 minute short-term memory, and 24 hour long-term memory, according to an modified protocol^98,259 as depicted in Figure 8A. The puzzle box test was conducted using the same apparatus used for the light-dark box test. For each trial in the puzzle box, mice were placed in the light side at the end farthest from the dark side and video recorded (C920, Logitech) until successfully passing through blockage and entering the dark side through the hole in the barrier. Mice were subjected to nine trials in total with three trials in immediate succession per day for three consecutive days. Blocking material was omitted in trial 1. Trial 2 through 9 had the floor elevated such that mice needed to climb down through a 4.2 x 4.2 cm hole in the floor to pass through the hole in the barrier. Trial 2, 3, and 4 had nothing blocking the hole. Trial 5, 6, and 7 had corncob bedding (Teklad 7097, Envigo) filling the hole. Trial 8 and 9 had crinkle cut paper strands (CNK, The Andersons) filling the hole. Mice were kept alone in a holding cage between trials on each day. The apparatus was cleaned with 70% ethanol between trials (3 minutes) and between mice. Videos were analyzed manually for a latency to enter the dark side. Higher latency was considered worse performance on the trial. As depicted in Figure 8A, performance was considered habituation to the apparatus in trial 1; problem solving ability in trial 2, 5, and 8; 3 minute memory during trial 3, 6, and 9; and 24 hour memory in trial 4 and 7. Latency to enter the dark side was logarithm transformed. Data for males and females were analyzed and graphed separately if there were sex-specific statistical effects.

Quantification and statistical analysis

Data were analyzed statistically and visualized using GraphPad Prism version 9.3.1 (471) (RRID:SCR_002798). Statistical comparisons for figures are described in each figure legend. GCMS metabolomic data (peak area intensities) were analyzed using principal component analysis and partial least squares discriminant-analysis using MetaboAnalyst 5.0 (RRID:SCR_015539)^260,261 without filtering. Data were log transformed and auto scaled. For all figures, ns = p>0.05, ∗ = p<0.05, ∗∗ = p<0.01, ∗∗∗ = p<0.001, and ∗∗∗∗ = p<0.0001. Data from males and females were consolidated wherever there were no sex differences detected.

Published: June 21, 2024