Targeting glucose-inhibited hippocampal CCK interneurons prevents cognitive impairment in diet-induced obesity

Summary

Metabolic disorders are closely linked to increased risk of cognitive decline, with Western-style high-fat diet (HFD) emerging as key contributors. However, the underlying cellular and molecular mechanisms remain unclear. Here, we demonstrate that short-term HFD (stHFD) consumption disrupts memory processing by inducing hyperactivity in dentate gyrus (DG) cholecystokinin-expressing interneurons (CCK-INs). We identify DG CCK-INs as glucose-inhibited neurons that become hyperactive in response to stHFD-induced reductions in DG glucose availability, coinciding with increased phosphorylation of the glycolytic enzyme pyruvate kinase M2 (PKM2). Restoring glucose availability, reducing PKM2 expression, or inhibiting PKM2 activity normalizes CCK-IN activity and rescues memory deficits. Furthermore, interventions preventing CCK-IN hyperactivity or PKM2 phosphorylation protect against long-term cognitive impairments in a diet-induced obesity mouse model. These findings reveal a previously unrecognized mechanism by which dietary metabolic stress disrupts hippocampal function and highlight DG CCK-INs and PKM2 as promising therapeutic targets for preventing cognitive decline associated with metabolic disorders.

Graphical Abstract

eTOC Blurb

Landry et al. show that short-term high-fat diet reduces hippocampal glucose availability, driving CCK-IN hyperactivity and memory deficits. Normalizing aberrant PKM2 or CCK-IN activity subsequently protects against obesity-related cognitive impairments. This study reveals a metabolic-neuronal mechanism linking diet to cognition and highlights potential targets for preventing diet-associated cognitive decline.

Introduction

The human brain requires a continuous and efficient supply of energy to sustain cognitive functions and daily activities. However, with minimal intrinsic energy reserves, it is highly dependent on nutrient availability from dietary sources. A growing body of evidence suggests that diet plays a crucial role in shaping cognitive function^1–10. In particular, the Western-style high-fat diet (HFD), rich in saturated fats, has emerged as a significant lifestyle factor contributing to both metabolic and cognitive dysfunction^1,3–8. Notably, even short-term HFD (stHFD) consumption has been linked to impaired cognitive performance in humans^11,12. Moreover, epidemiological studies indicate that individuals with metabolic syndrome face an elevated risk of developing cognitive impairments and neurodegenerative diseases such as Alzheimer’s disease¹³. Thus, understanding the mechanisms linking metabolic dysfunction to cognitive decline is essential for identifying early intervention strategies to mitigate these risks.

While numerous pathological features of HFD consumption have been identified, including insulin resistance^3,14–17, oxidative stress¹⁸, inflammation^6,7,14,19, transcriptional dysregulation^7,20, and synaptic plasticity deficits^3,6,7, the precise cellular and molecular mechanisms underlying HFD-induced cognitive impairment remain poorly understood. In particular, brain region- and cell type-specific alterations in neuronal dynamics during the early stages of HFD consumption remain largely unexplored. Identifying the most vulnerable neuronal populations and their molecular alterations is critical for developing strategies to prevent cognitive decline associated with metabolic disorders such as obesity and type 2 diabetes.

To address these questions, we systematically record in vivo calcium dynamics as a proxy for neuronal activity in major neuronal populations of the dentate gyrus (DG), a key hippocampal region involved in memory processing, to identify early neural alterations induced by stHFD. We find that stHFD consumption leads to memory deficits driven by hyperactivity of DG cholecystokinin-expressing interneurons (CCK-INs). Mechanistically, we identify DG CCK-INs as a previously unrecognized class of glucose-inhibited neurons that become hyperactive under stHFD conditions due to reduced DG glucose uptake and increased phosphorylation of the glycolytic enzyme pyruvate kinase M2 (PKM2). Restoring glucose levels or reducing PKM2 activity via genetic, pharmacological, and dietary interventions rescues these deficits. Finally, leveraging these mechanistic insights, we demonstrate that early interventions targeting CCK-IN hyperactivity or PKM2 phosphorylation prevent cognitive decline in a long-term diet-induced obesity (DIO) model.

Results

Hippocampal interneuron hyperactivity mediates stHFD-induced cognitive deficits

Recent studies indicate that even brief exposure to a HFD can impair cognitive function, with the hippocampus being particularly vulnerable to HFD-associated damage^4–7. To investigate these effects, we assessed hippocampus-dependent behaviors using a battery of tests established in our previous studies^21,22. Wild-type (WT) mice (8–9 weeks old) were placed on a HFD (58% fat, 25% carbohydrate, 17% protein) for two days before beginning behavioral testing (Figure S1A). As expected, HFD-fed mice exhibited increased caloric intake; however, body weight and blood glucose levels remained unchanged compared to chow-fed controls throughout the behavioral testing period (Figure S1B–D). Notably, HFD-fed mice displayed significant impairments in hippocampus-dependent spatial and contextual memory, as evidenced by a reduced discrimination ratio in the novel place recognition (NPR) test and decreased freezing time in the contextual fear conditioning (CFC) test (Figure S1G, S1I). Additionally, HFD-fed mice exhibited increased locomotion in the open field test (Figure S1E). In contrast, measures of anxiety- and depressive-like behaviors remained unaffected, including time spent in the center of the open field, time in the open arms of the zero maze, and immobility time in the forced swim test (Figure S1F, S1H, S1J). These findings suggest that stHFD selectively disrupts hippocampus-dependent cognitive functions without significantly altering affective behaviors.

To investigate whether hippocampal circuit dynamics are altered by stHFD exposure, we focused on the dentate gyrus (DG), the primary input region of the hippocampal trisynaptic circuit. Using fiber photometry, we recorded population calcium dynamics to assess baseline activity in two major DG neuronal populations, granule cells (GCs) and interneurons (IN’s), in chow-fed control mice (chow mice) and mice fed HFD for five days (stHFD mice). AAVs expressing the calcium indicator GCaMP were delivered to the DG of WT or VGAT-Cre mice to selectively label GCs or interneurons, respectively. Notably, baseline activity of DG IN’s was significantly elevated in stHFD mice, whereas GC activity remained unchanged (Figure 1A–E, Figure S1K–O).

Figure 1. DG interneuron hyperactivity mediates stHFD-induced cognitive deficits.

(A) Experimental approach (white box indicates optical fiber path), (B) Representative photometry traces, (C) Frequency of spikes, (D) Amplitude of spikes, and (E) AUC of basal interneuron GCAMP in chow v. stHFD mice (two-tailed, unpaired t-tests; C, p=0.0352; D, p=0.0838; E, p=0.0386; n=13–17 mice). Spikes in GCAMP defined as 3 standard deviations (3S.D.) above the mean. (F) Timeline for chemogenetic manipulation of DG interneurons. (G-H) Effects of chemogenetic activation of DG interneurons 1.5h before NPR encoding in chow mice (two-tailed, unpaired t-test; p=0.0122; n=6–8 mice). (I-J) Effects of chemogenetic inbibition of DG interneurons 1.5h before NPR encoding in stHFD mice (two-tailed, unpaired t-test; p=0.0131; n=10–13 mice). Data presented as mean±S.E.M * indicates p<0.05. Scale bars = 200 um.

To determine whether IN hyperactivity contributes to stHFD-induced cognitive deficits, we chemogenetically activated DG INs by delivering Cre-dependent AAVs expressing excitatory hM3Dq into VGAT-Cre mice, followed by intraperitoneal (IP) injection of clozapine N-oxide (CNO) 1.5 h before the encoding phase of the NPR test (Figure 1F–G), a well-established DG-dependent memory task^23–25. Strikingly, acute IN activation impaired NPR performance in otherwise healthy chow mice (Figure 1H), suggesting that excessive DG IN activity disrupts cognitive function. Conversely, we tested whether inhibiting DG INs could rescue stHFD-induced cognitive deficits. Cre-dependent AAVs expressing inhibitory hM4Di were delivered into VGAT-Cre stHFD mice to selectively suppress IN activity before encoding (Figure 1F, 1I). Indeed, acute IN inhibition significantly improved NPR performance in stHFD mice (Figure 1J). Importantly, food intake remained unchanged in all experiments, ruling out confounding metabolic factors (Figure S2A–B). Together, these findings demonstrate that DG interneuron hyperactivity mediates the cognitive deficits induced by stHFD, highlighting a critical circuit mechanism underlying diet-driven memory impairments.

Hyperactivity of dentate cholecystokinin-expressing interneurons mediates stHFD-induced cognitive deficits

To identify the IN subpopulation responsible for stHFD-induced cognitive deficits, we selectively inhibited parvalbumin- (PV), neuropeptide Y- (NPY), or cholecystokinin-expressing (CCK) INs in the DG by delivering Cre-dependent AAVs expressing hM4Di into PV-Cre, NPY-Cre, or CCK-Cre stHFD mice (Figure 2A & Figure S2C–I). Remarkably, inhibition of DG CCK interneurons (CCK-INs), but not PV or NPY INs, 1.5h prior to encoding significantly improved memory performance in the NPR test in stHFD mice (Figure 2B & Figure S2C–I). Supporting these findings, cFOS staining and fiber photometry revealed elevated baseline activity of DG CCK-INs in stHFD mice (Figure 2C–I). To further examine CCK-IN activity during cognitive performance, we recorded calcium dynamics during the NPR test. stHFD mice exhibited heightened calcium responses in CCK-INs while exploring objects and locations during both the encoding and retrieval phases of the NPR task (Figure 2J–S & Figure S2K–L), suggesting that CCK-IN hyperactivity may interfere with task-specific information processing. To directly test this hypothesis, we mimicked the hyperactive CCK-IN state observed in stHFD mice by acutely activating CCK-INs with hM3Dq in chow mice. This manipulation significantly impaired memory performance (Figure 2T–U), further supporting the role of CCK-IN hyperactivity in cognitive dysfunction. Notably, food intake remained unchanged across experiments (Figure S2C–J), ruling out confounding metabolic factors. Together, these findings demonstrate that DG CCK-IN hyperactivity mediates stHFD-induced cognitive deficits and that suppressing CCK-IN activity rescues memory impairments associated with stHFD.

Figure 2. Hyperactivity of dentate CCK-INs mediates stHFD-induced cognitive deficits.

(A) Experimental approach for chemogenetic CCK-IN inhibition before (B) NPR testing in stHFD mice (two-tailed, unpaired t-test; p=0.0001; n=7–9 mice). (C) Representative images and (D) Quantification of basal CCK-IN cFOS expression in chow v. stHFD mice (two-tailed, unpaired t-test; p=0.0277 n=9–11 mice) (arrows indicate examples of cFOS and CCK-IN colocalization). (E) Experimental approach, (F) Representative photometry traces, (G) Frequency of spikes, (H) Amplitude of spikes, and (I) AUC of basal CCK-IN GCaMP in chow v. stHFD mice (two-tailed, unpaired t-tests; G, p=0.0006; H, p=0.4656; I, p=0.4849; n=10–15 mice). Spikes in GCaMP defined as 3 standard deviations (3S.D.) above the mean. (J-L) Perievent CCK-IN GCaMP analysis when investigating objects during NPR encoding (M) Peak GCaMP amplitude and (N) AUC (two-tailed, unpaired t-tests; M, p=0.0155, N, p=0.2353; n=74–118 events from 8–9 mice). (O-S) Perievent CCK-IN GCaMP analysis during NPR retrieval (R, p=0.0113, S, p=0.1398; two-tailed unpaired t-tests; n=66–99 events from 8–9 mice). (T) Experimental approach for chemogenetic activation of CCK-INs before (U) NPR testing in chow mice (two-tailed, unpaired t-test; p=0.0434; n=11 mice/group). Scale bars = 200um for A, E, and T; scale bar = 100um for C. Data presented as mean±S.E.M. * indicates p<0.05, *** = p<0.001, **** = p<0.0001

CCK-IN activity is inversely correlated with glucose availability

A recent study reported that GLUT1, a glucose transporter predominantly expressed in vascular endothelial cells (ECs) at the blood–brain barrier, is downregulated across the brain, including the hippocampus, after just 3 days of HFD exposure²⁶. This resulted in brain-wide reductions in glucose uptake and neuronal glucose availability due to the critical role of GLUT1 in endothelial cell-mediated diffusion of glucose to neurons^26–28. Considering glucose’s role as the primary energy source to neurons, reduced availability would have significant implications regarding neuronal activity and function^29–31. To obtain region-specific insights, we first confirmed selective expression of GLUT1 in DG ECs by reanalyzing a published single-nucleus RNA sequencing (snRNA-Seq) dataset³² (Figure 3A). We then validated a significant reduction in GLUT1 expression in the DG of stHFD mice using immunohistochemistry (Figures 3B–C). Furthermore, the average distance between CCK-INs and the nearest GLUT1+ ECs was significantly increased in stHFD mice (Figure 3D), indicating impaired glucose diffusion to CCK-INs. Together, these findings indicate that reduced glucose uptake and impaired glucose diffusion are likely to contribute to reduced glucose availability to CCK-INs.

Figure 3. CCK-IN activity is inversely correlated with glucose availability.

(A) Violin plot showing normalized Slc2a1 (GLUT1) expression of 40403 dentate gyrus cells from Langlieb, et. al, 2023⁶¹, demonstrating selective expression in endothelial cells. (B) Representative images and quantification of (C) Total DG GLUT1 expression and (D) Average distance between CCK-INs and nearest GLUT⁺ cell in chow v. stHFD mice (two-tailed, unpaired t-tests; C, p=0.0017, D, p<0.0001; n=12–14 mice and 184–266 CCK-INs). (E) Representative GC GCaMP ROI’s in ex vivo live brain slices undergoing high (10mM) to low (1mM) glucose concentrations measured by 2-photon microscopy. (F) Frequency, (G) Amplitude, and (H) AUC of GCaMP spikes (determined by 2 standard deviations above the mean) (two-tailed, unpaired t-tests; F, p=0.0206, G, p=0.0555, H, p=0.0117; n=51 ROI’s from 9 slices from 4 mice. (I) Representative CCK-IN GCaMP ROI’s in ex vivo live brain slices undergoing high (10mM) to low (1mM) glucose concentrations. (J) Frequency, (K) Amplitude, and (L) AUC of GCaMP spikes (two-tailed, unpaired t-tests; J, p<0.0001, K, p=0.0340, L, p<0.0001; n=85 ROI’s from 10 slices from 3 mice). (M-P) Effects of a single IP glucose injection (2g/kgBW) on CCK-IN GCaMP in chow mice (One-way repeated measures ANOVA, with Dunnett’s correction for comparisons to baseline; N, 10min-p=0.2022, 20min-p=0.0410, 30min-p=0.0227, 60min-p=0.0513, 90min-p=0.7730, 120min-p=0.4100; O, 10min-p=0.8649, 20min-p=0.05780, 30min-p=0.3156, 60min-p=0.5001, 90min-p=0.5220, 120min-p=0.7686; P, 10min-p=0.1910, 20min-p=0.0453, 30min-p=0.0200, 60min-p=0.0397, 90min-p=0.4650, 120min-p=0.2390; n=8 mice). (Q-T) Effects of refeeding after an overnight fast on CCK-IN GCaMP in chow mice (One-way repeated measures ANOVA, with Dunnett’s correction for comparisons to baseline; R, 10min-p=0.0539, 20min-p=0.0230, 30min-p=0.0204, 60min-p=0.0230, 90min-p=0.0294, 120min-p=0.0137; S, 10min-p>0.9999, 20min-p=0.9990, 30min-p>0.9999, 60min-p=0.9998, 90min-p=0.3580, 120min-p=0.4021; T, 10min-p=0.1984, 20min-p=0.0278, 30min-p=0.0407, 60min-p=0.0545, 90min-p=0.0324, 120min-p=0.0292; n=6 mice). Scale bar = 50uM. Data presented as mean±S.E.M. * indicates p<0.05, ** = p<0.01, **** = p<0.0001.

To date, the implications of changes in glucose availability on distinct DG neuronal populations remain largely unexplored. To determine whether DG neurons, particularly CCK-INs, are glucose-sensitive, we performed two-photon (2P) calcium imaging in ex vivo acute brain slices from chow-fed mice expressing GCaMP in CCK-INs, granule cells (GCs), and VGAT+ INs. We applied a glucose step from high (10 mM) to low (1 mM) in artificial cerebrospinal fluid (aCSF) to mimic reduced glucose availability induced by HFD²⁶. Interestingly, a significant reduction in GC activity was observed in response to reduced glucose concentrations, indicating that GCs are glucose-excited (Figure 3E–H). In contrast, VGAT+ INs and CCK-INs showed increased activity in response to reduced glucose levels (Figure 3I–L, S3A–D), indicating that DG CCK-INs are glucose-inhibited. Considering these findings we hypothesized that hyperactivity of DG CCK-INs observed in stHFD mice may be driven by reduced glucose availability associated with impaired brain glucose uptake and decreased GLUT1 expression.

To confirm that DG CCK-IN activity is inversely correlated with glucose availability in vivo, we fasted healthy chow-fed mice overnight to lower systemic glucose levels and measured GCaMP signals before and after acute glucose injection or refeeding for 2 hours. Consistent with our ex vivo findings, glucose injection led to a marked reduction in CCK-IN activity within 20–60 minutes (Figure 3M–P). Ad libitum refeeding similarly suppressed CCK-IN activity within 20 minutes and the effects persisted for more than 2 hours, likely due to repeated bouts of feeding during recording (Figure 3Q–3T). Notably, these inhibitory effects of glucose and refeeding were also observed in VGAT+ INs, albeit with slightly different response dynamics due to heterogeneous populations of INs labeled in the VGAT-Cre line (Figure S3M–T), while GCs exhibited the opposite pattern, as indicated by increased cFOS expression and GCaMP baseline shifts (Figure S3E–L). Together, our ex vivo and in vivo findings identify novel glucose-sensitive neuronal populations in the DG, in which GCs function as glucose-excited cells and CCK-INs (and DG interneurons more broadly) are glucose-inhibited. These results provide critical insights into glucose-mediated regulation of DG circuit dynamics and how dietary influences that alter glucose availability could have major implications regarding cognitive function.

Increased glucose availability rescues CCK-IN hyperactivity and cognitive deficits in stHFD mice

Considering evidence that brain glucose uptake is reduced in stHFD mice²⁶ and previous studies demonstrating the cognitive benefits of glucose administration during memory tasks^33–35, we aimed to test the hypothesis that increasing glucose delivery mitigates cognitive deficits in stHFD mice by assessing NPR task performance following (IP) saline or high-dose (2g/kgBW) glucose injection one hour before memory encoding. As an alternative method of improving glucose delivery, we also determined the effects of acute feeding on memory performance by comparing ad libitum-fed stHFD mice to those fasted overnight and refed with HFD one hour before memory encoding. Similar to chow-fed controls, both IP glucose and refeeding inhibited CCK-IN activity in stHFD mice (Figure 4A–H), which strikingly coincided with improved memory performance (Figure 4I–K).

Figure 4. Increased glucose availability rescues CCK-IN hyperactivity and cognitive deficits in stHFD mice.

(A-D) Effects of a single IP glucose injection (2g/kgBW) on CCK-IN GCaMP in stHFD mice (One-way repeated measures ANOVA, with Dunnett’s correction for comparisons to baseline; B, 10min-p=0.0118, 20min-p=0.0126, 30min-p=0.0093, 60min-p=0.0257, 90min-p=0.0028, 120min-p=0.0071; C, 10min-p=0.6833, 20min-p=0.3238, 30min-p=0.9411, 60min-p=0.8846, 90min-p>0.9999, 120min-p=0.7718; D, 10min-p=0.3323, 20min-p=0.1234, 30min-p=0.6215, 60min-p=0.9622, 90min-p=0.4318, 120min-p=0.7316; n=6 mice. (E-H) Effects of refeeding after an overnight fast on CCK-IN GCaMP in stHFD mice (One-way repeated measures ANOVA, with Dunnett’s correction for comparisons to baseline; F, 10min-p=0.9996, 20min-p=0.0358, 30min-p=0.0214, 60min-p=0.0016, 90min-p=0.0020, 120min-p=0.0010; G, 10min-p=0.8712, 20min-p=0.9197, 30min-p=0.8813, 60min-p=0.8049, 90min-p=0.4716, 120min-p=0.3548; H, 10min-p=0.9917, 20min-p=0.1349, 30min-p=0.0350, 60min-p=0.0129, 90min-p=0.0159, 120min-p=0.0166; n=8 mice. (I) Experimental timeline for the effects of (J) a single IP glucose injection (2g/kgBW) or (K) refeeding after an overnight fast on NPR performance (glucose injection and refeeding performed 1h. prior to encoding phase) (two-tailed, unpaired t-tests, J, p=0.0090, n=8–9 mice/group; K, p=0.0270, n=10–13 mice/group). (L-M) Experimental design and timeline investigating the effects of chemogenetic CCK-IN activation on (N) Glucose- and (O) Fast/refeed-associated rescue of NPR performance (one-way ANOVA’s with Tukey’s correction for multiple comparisons; N, saline mCherry v. glucose mCherry-p=0.0260; glucose mCherry v. glucose hM3-p=0.0062; n=15–16 mice/group; O, ad libitum fed v. refeed mCh-p=0.0225; refeed mCh v. refeed hM3-p=0.0422; n=9–10 mice/group). Data presented as mean±S.E.M. * indicates p<0.05, ** = p<0.01.

To confirm that increased glucose availability rescues memory deficits in stHFD mice via CCK-IN inhibition, we used hM3Dq to acutely activate CCK-INs prior to glucose injections or refeeding before memory encoding (Figure 4L–O). CCK-IN activation abolished the memory-enhancing effects of both glucose injection and refeeding (Figure 4N–O), highlighting CCK-IN inhibition as a key mechanism by which improved glucose delivery restores cognitive function. Together, these findings provide direct evidence that acute interventions increasing glucose availability mitigate CCK-IN hyperactivity and improve memory performance in stHFD mice.

Increased phosphorylation of PKM2^ser37 links CCK-IN hyperactivity to cognitive deficits in stHFD mice

The glycolytic enzyme pyruvate kinase M2 (PKM2) is often upregulated in response to low glucose availability^36–38. When phosphorylated at serine 37^39,40, PKM2 translocates to the nucleus, driving transcriptional changes that promote aerobic glycolysis, a phenomenon known as the Warburg effect, originally described in tumor cells^41–45. In line with this, we observed elevated total PKM2 and, more importantly, phosphorylated PKM2 at serine 37 (pPKM2^ser37) in DG CCK-INs, with no significant changes in GCs, suggesting that stHFD selectively enhances PKM2 activity in DG CCK-INs (Figure 5A–B, S4A–D). Moreover, CCK-INs of stHFD mice exhibited elevated phosphorylation of threonine 11 on histone 3 (pH3^t11) (Figure 5C–D), a direct downstream target of nuclear PKM2, along with reduced expression of PKM1 (Figure S4E–F), the isoform that promotes oxidative phosphorylation. These findings suggest a shift from homeostatic PKM1 activity toward nuclear PKM2 signaling in response to stHFD. Interestingly, pPKM2^ser37 expression in stHFD mice was significantly reduced 1 hour after IP glucose injection or refeeding after an overnight fast (Figure 5E–H), highlighting improved glucose delivery as a method of normalizing pPKM2^ser37 levels.

Figure 5. Increased PKM2 activity links CCK-IN hyperactivity to cognitive deficits in stHFD mice.

(A) Representative images and quantification of pPKM2^ser37 expression in (B) CCK-INs of chow v. stHFD mice (two-tailed, unpaired t-test, p=0.0042; n=5–6 mice). (C) Representative image and (D) Quantification of threonine 11 phosphorylation on histone 3 (pH3^t11) in Chow v. stHFD mice (two-tailed, unpaired t-test; p=0.0036; n=8–11 mice/group). (E) Representative images and (F) Quantification of pPKM2^ser37 expression in CCK-INs 1h. after acute IP glucose injection (2g/kgBW) in stHFD mice (two-tailed, unpaired T-test; p=0.0001; n=7 mice/group). (G) Representative images and (H) Quantification of cFOS expression in CCK-INs in ad libitum v. fast/refed stHFD mice (two-tailed, unpaired t-test; p=0.0277; n=9–11 mice) (white arrows indicate colocalization). (I) Experimental approach for shRNA-medated PKM2 knockdown specifically in CCK-INs. Effects of shPKM2 on (J-K) CCK-IN pPKM2^ser37 expression (n=6–8 mice) and (L) NPR performance (n=10–11 stHFD mice) (two-tailed, unpaired t-tests; K, p=0.0166; L, p=0.0421). (M) Representative images and (N) Quantification of basal cFOS expression in CCK-INs of stHFD mice with scRNA or shPKM2 (two-tailed, unpaired t-tests; p=0.0062; n=5–7 mice) (white arrows indicate examples of pPKM2^ser37 and CCK-IN colocalization). (O) Representative traces, (P) Frequency of spikes, (Q) Amplitude of spikes, and (R) AUC of spikes in basal CCK-IN GCaMP in stHFD mice with scRNA or shPKM2 (two-tailed, unpaired t-tests; P, p=0.0336; Q, p=0.8763; R, p=0.0849; n=9–12 mice). Spikes in GCaMP defined as 3 standard deviations (3S.D.) above the mean. (S) Experimental timeline for effects of ICV shikonin on (T) Food intake (note Veh PF food intake data is perfectly covered by shikonin data) and (U) NPR performance in stHFD mice (Veh = vehicle; PF = pair-fed). (T, two-way ANOVA, vehicle v. shikonin-p=0.0224; vehicle v. veh PF-p=0.057; shikonin v. veh PF-p>0.9999); U, one way ANOVA with Tukey’s correction for multiple comparisons; vehicle v. shikonin-p=0.0028; vehicle v. veh PF-p=0.1869; shikonin v. veh PF-p<0.0001; n=9–16 mice). Data presented as mean±S.E.M. * indicates p<0.05, ** = p<0.01, **** = p<0.0001. Scale bars = 100um for A, C, E, G, J, and M; scale bar = 200um for I.

To determine the causal role of PKM2 in stHFD-induced cognitive deficits, we used Cre-dependent AAVs expressing a validated shRNA against PKM2 (shPKM2) to selectively knock down PKM2 in CCK-INs⁴⁶ (Figure 5I–K). Knockdown efficiency was confirmed by a significant reduction in the percentage of CCK-INs expressing PKM2 and pPKM2^ser37 compared with scrambled shRNA (scRNA) controls (Figure 5J–K, S4G–H). Strikingly, stHFD mice with CCK-IN-specific PKM2 knockdown exhibited improved memory performance in the NPR test compared with scRNA controls (Figure 5L), which was accompanied by reduced CCK-IN activity, evidenced by fiber photometry and cFOS expression (Figure 5M–R), despite no changes in food intake or body weight (Figure S4I–J).

To further validate the memory-rescuing effects of PKM2 inhibition, we pharmacologically targeted PKM2 nuclear translocation using shikonin, a well-established PKM2 inhibitor⁴⁶. However, daily intracerebroventricular (ICV) shikonin treatment unexpectedly reduced food intake and caused weight loss in stHFD mice, necessitating the use of a pair-feeding model (Figure 5S–T, S4K–L). Pair-fed mice received the same amount of food as their shikonin-treated counterparts to control for differences in caloric intake. Despite normalizing for reduced food intake, shikonin-treated stHFD mice exhibited improved memory in the NPR test compared to both vehicle-treated ad libitum-fed stHFD mice and pair-fed stHFD controls (Figure 5U). Overall, these findings suggest that stHFD increases PKM2 expression and nuclear activity in DG CCK-INs, contributing to CCK-IN hyperactivity and cognitive deficits.

Early interventions targeting CCK-IN abnormalities prevent cognitive decline in diet-induced obesity mice

Leveraging our novel discoveries in stHFD mice, we investigated whether targeting CCK-IN abnormalities could prevent cognitive deficits in a chronic metabolic disorder model by using the well-established diet-induced obesity (DIO) mouse model of type 2 diabetes⁴⁷. DIO mice were fed HFD for 10 weeks, becoming overweight by week 7 (Figure 6A). These mice exhibited a reduced discrimination ratio in the NPR task and spent less time in the center of the open field, indicative of impaired spatial memory and a mild anxiety phenotype. Notably, locomotion and performance in the zero maze, contextual fear conditioning (CFC), and forced swim tests remained unchanged (Figure 6B–C & Figure S5A–D). Strikingly, chronic inhibition of DG CCK-INs via hM4Di and CNO in drinking water or long-term PKM2 knockdown in DG CCK-INs rescued spatial memory deficits in DIO mice, as evidenced by improved NPR performance (Figure 6D–J). However, neither intervention altered time spent in the center of the open field test (Figure 6F, 6K), suggesting a selective role of DG CCK-INs and PKM2 on cognitive function rather than anxiety-related behavior. These findings underscore the potential of early interventions targeting neuronal abnormalities induced by metabolic insults to mitigate long-term cognitive decline in metabolic diseases.

Figure 6. Interventions targeting early HFD-associated neuronal abnormalities prevent cognitive decline in DIO mice.

(A) Time course of body weight in the DIO mouse model of Type-2 diabetes (two-way repeated measures ANOVA with Bonferroni’s correction for multiple comparisons; Group effects at week 7-p=0.0219, week 8-p=0.0311, week 9-p=0.0177, and week 10-p=0.0031; n=19–31 mice/group). Performance of DIO mice in the (B) NPR task and (C) Open field task: time in center (two-tailed, unpaired t-tests; B, p=0.0045, C, p=0.0014; n=8–11 mice/group). (D) Experimental timeline for effects of long term chemogenetic inhibition of CCK-INs via 2mg/kg CNO water in DIO mice. Effects of long term hM4 in CCK-INs on (E) NPR and (F) Open Field tasks in DIO mice (two-tailed, unpaired t-tests; E, p=0.0394, F, p=0.2158; n=8–16 mice/group). (G) Experimental timeline for effects of longterm shPKM2 on behavior in DIO mice. (H-I) Validation of pPKM2 knockdown, (J) NPR and (K) Open Field tasks (two-tailed, unpaired t-tests; I, p=0.0221 J p=0.0117, K, p=0.6487; n=6–14 mice/group). Data presented as mean±S.E.M. * indicates p<0.05, ** = p<0.01, *** = p<0.001. Scale bar = 100um.

Discussion

The increasing prevalence of metabolic disorders, particularly obesity and type 2 diabetes, has raised significant concerns regarding their impact on cognitive function. While numerous studies have established a strong association between HFD consumption and cognitive decline^48,49, the precise cellular and molecular mechanisms underlying these impairments remain poorly understood. Our study identifies a novel mechanism by which HFD-induced metabolic stress impairs cognitive function through hyperactivity of glucose-inhibited DG CCK-INs and PKM2-mediated metabolic reprogramming. By demonstrating that early metabolic interventions can prevent cognitive decline, our findings highlight new therapeutic avenues for addressing the growing burden of obesity-related cognitive decline.

Selective vulnerability of DG CCK-INs to HFD-induced metabolic insults

Our findings suggest that DG CCK-INs are particularly vulnerable to metabolic stress induced by stHFD consumption. We observed reduced GLUT1 expression in the DG of stHFD mice, and further showed that glucose spikes suppress pPKM2 expression and CCK-IN hyperactivity, thereby improving memory performance. Together, these results indicate that stHFD reduces GLUT1 levels in the DG, which is associated with diminished brain glucose uptake and impaired glucose delivery to DG neurons. Despite the well-established role of glucose in neuronal function, the impact of glucose availability on specific DG neuronal subtypes remains largely unexplored. Previous studies have classified glucose-sensitive neurons into two categories, glucose-excited and glucose-inhibited, primarily in the hypothalamus, where they regulate glucose homeostasis⁵⁰. For instance, proopiomelanocortin (POMC) neurons in the arcuate nucleus (ARC), which promote satiety, are glucose-excited and increase their activity in response to rising glucose levels, primarily through ATP-sensitive potassium (KATP) channel closure^31,51. Conversely, hunger-inducing neuropeptide Y (NPY) neurons in the ARC and lateral hypothalamus (LH) are glucose-inhibited^52,53, increasing their activity under low-glucose conditions via mechanisms such as leak K⁺ channels^54,55 and AMPK-mediated neuronal nitric oxide synthase (nNOS) activation, which leads to Cl⁻ channel closing⁵⁶. Extending this framework to hippocampal interneurons, our study identifies DG CCK-INs as glucose-inhibited neurons that become hyperactive when glucose levels decline following stHFD consumption. In contrast, DG GCs are glucose-excited neurons whose activity decreases when glucose levels are reduced. This finding underscores the metabolic sensitivity of hippocampal circuits and suggests that distinct neuronal subtypes within the DG may differentially respond to dietary-induced metabolic shifts. The selective susceptibility of CCK-INs to HFD-induced disruptions in glucose homeostasis highlights a previously unrecognized mechanism linking metabolic stress to hippocampal dysfunction and cognitive decline.

PKM2 as a metabolic regulator of neuronal dysfunction

Our study identifies PKM2 as a critical molecular mediator linking HFD-induced metabolic stress to neuronal dysfunction and cognitive impairment. PKM2, a key glycolytic enzyme, exists in both tetrameric and dimeric forms, with its activity being modulated by post-translational modifications such as phosphorylation at Serine 37. Under pathological conditions, increased PKM2 phosphorylation leads to metabolic reprogramming, a phenomenon often referred to as the Warburg effect, which shifts cellular metabolism from oxidative phosphorylation to glycolysis⁵⁷. While this metabolic adaptation has been widely studied in cancer⁵⁸, emerging evidence suggests that similar mechanisms may contribute to neurological disorders, including Alzheimer’s disease (AD), where upregulated PKM2 disrupts neuronal metabolism and pharmacological inhibition of PKM2 restores normal metabolic function in AD patient-derived neurons⁵⁹. Our findings provide new evidence that PKM2 phosphorylation in DG CCK-INs plays a causal role in HFD-induced cognitive decline. This highlights PKM2-mediated glycolysis as a shared pathological mechanism across various conditions involving impaired glucose metabolism. The observation that PKM2 inhibition restores CCK-IN function and rescues memory deficits underscores its potential as a therapeutic target for mitigating metabolic disorder-associated cognitive impairments.

Therapeutic potential of early interventions targeting cognitive and metabolic dysfunction

A key translational aspect of our study is the demonstration that early interventions targeting CCK-IN hyperactivity or PKM2 phosphorylation can prevent long term cognitive decline in diet-induced obesity. This suggests that metabolic interventions, including dietary modifications or pharmacological approaches, may be effective in preserving cognitive function in individuals at risk for obesity-related neurodegeneration. Notably, we found that dietary interventions incorporating acute fasting periods following HFD consumption were sufficient to normalize DG CCK-IN activity and improve memory function, further supporting the idea that metabolic stress on the brain can be mitigated through strategic nutritional strategies.

Additionally, our findings with the PKM2 inhibitor shikonin suggest a dual role for this compound in both cognitive and metabolic improvements. While shikonin restored memory function in stHFD mice by targeting DG CCK-INs, it also reduced food intake and induced weight loss, indicating that its metabolic effects may extend beyond the hippocampus to brain regions involved in feeding regulation. This suggests that PKM2 modulation may offer a converging therapeutic approach for simultaneously addressing both cognitive and metabolic dysfunctions associated with obesity.

Implications and future directions

Our study provides a mechanistic framework for understanding how HFD consumption disrupts hippocampal function at the cellular and molecular levels. However, several important questions remain. First, while our data suggest that DG CCK-INs are selectively vulnerable to HFD-induced metabolic stress when comparing to PV- and NPY-interneurons, it is possible that other interneuron subtypes also contribute to diet-induced cognitive impairment. Future studies should explore the extent to which other abundant DG interneurons, such as somatostatin interneurons, may be affected by HFD consumption. Second, the precise downstream effects of CCK-IN hyperactivity on hippocampal network dynamics and memory processing remain to be fully elucidated. Given that CCK-INs play key roles in regulating excitatory-inhibitory balance, their dysregulation may have widespread effects on hippocampal oscillations and synaptic plasticity⁶⁰. Electrophysiological and in vivo imaging studies could provide further insights into how metabolic disruptions alter network function in the hippocampus. Finally, our findings raise the intriguing possibility that PKM2-mediated metabolic reprogramming may be a common mechanism underlying cognitive deficits across various neurological disorders. Given its involvement in AD and other neurodegenerative diseases, future research should investigate whether targeting PKM2 can provide therapeutic benefits beyond metabolic disorder-associated cognitive impairment.

In summary, our study identifies a novel mechanism by which HFD-induced metabolic stress impairs cognitive function through hyperactivity of glucose-inhibited DG CCK-INs and PKM2-mediated metabolic reprogramming. By demonstrating that early metabolic interventions can prevent cognitive decline, our findings highlight new therapeutic avenues for addressing the growing burden of obesity-related neurodegeneration. As metabolic disorders continue to rise globally, understanding how dietary factors influence brain function will be crucial for developing effective strategies to preserve cognitive health.

RESOURCE AVAILABILITY

Lead Contact

Information and requests should be directed to and will be fulfilled by the Lead Contact, Juan Song (juansong@email.unc.edu).

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

• Code is available in a repository at https://doi.org/10.5281/zenodo.16703985.⁶¹

• No new snRNA-seq datasets were generated in this paper. Single nuclei counts, cell metadata, and library metadata from Langlieb, et. al, 2023⁶² were downloaded from https://docs.braincelldata.org/downloads.

• Any original data reported in this paper or additional information required to reanalyze the data are available from the lead contact upon request.

STAR★Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Adult male and female mice (8–9 weeks) with the C57BL6 background were used in this study and sex was matched between groups. C57BL/6J, CCK-Cre (Cat#: 012706), NPY-Cre (Cat#: 027851), PV-Cre (Cat#: 017320), and VGAT-Cre (Cat#: 016962) mice were obtained from Jackson laboratory, and all transgenic mice used were heterozygous. Mice were cared for in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and experimental protocols were approved by Institutional Animal Care and Use Committees of University of North Carolina Chapel Hill. Mice were housed at 20–22°C with a 12-h light-dark cycle. For stereotaxic injections, 5–6-week-old male and female mice were used, and 3 weeks were given for recovery before experiments. For diet-induced obesity mice, dietary intervention began at 8–9 weeks of age and continued for 10 weeks.

METHOD DETAILS

Dietary Manipulations

Chow mice were fed a standard chow diet (15% fat, 62% carbohydrate, 23% protein) while stHFD and DIO mice were fed a Western-style, saturated fat-rich diet (58% fat, 25% carbohydrate, 17% protein; D12331; Research Diets, New Brunswick, NJ)⁴⁷. stHFD mice were fed HFD for 5 days for all experiments, in accordance with the timepoint at which detriments in NPR performance were observed (Fig. S1G), while DIO mice were fed HFD for 10 weeks before experiments. Food intake was measured by weighing remaining food, subtracting from the total food provided, and normalizing to body weight. Bedding was inspected thoroughly for residual bits of food, which were included in measurements. For refeeding experiments, mice were subjected to an overnight fast, followed by ad libitum refeeding with the appropriate diet. For pair-feeding experiments, the pair-fed group was provided the same amount of food (normalized to body weight) as was consumed by the ad libitum treatment group.

Stereotaxic Injections

Stereotaxic injections were performed on animals 5–6 weeks of age as previously described⁶³. Mice recovered for 3 weeks after surgery before experiments. Mice were initially anesthetized under 4% isoflurane in oxygen at 1.0 LPM flow rate and maintained under 1.5% isoflurane. An incision of ~1cm was performed along the midline of the head and holes drilled in the skull at the desired coordinates. Virus was injected by microsyringe (Hamilton, 33GA) and microinjection pump (Harvard Apparatus), at a rate of 75 nl/min with the following coordinates from bregma: Dorsal DG GC’s (upper blade) - 1.9 mm posterior, ±1.5 mm lateral, 2.07 mm deep, dorsal hilus – 1.9 mm posterior, ±1.5 mm lateral, 2.1 mm deep. Injections for photometry experiments were performed unilaterally, while all other injections were bilateral. A total of 150–300 nl of virus was delivered to each site and the needle was left in the site for at least 10 min to permit diffusion. For ICV cannulations, cannulas were placed 1.0 mm lateral, 0.5 mm posterior, and 2.5 mm deep. For optical fiber implantation (Newdoon Inc, O.D.: 1.25 mm, core: 200 μm, NA: 0.37), fibers were placed 0.01 mm above viral injection. Cannulas and optical fibers were fixed in place with dental cement (Stoelting, Wood Dale, IL, USA).

Behavior Testing

All behavior experiments were performed as described previously^63,64, in a dim lit room (~60 lux) with 40 dB background noise. After all trials, the apparatuses were cleaned with 70% ethanol (and water replaced during the Forced Swim test). For all analyses, unless otherwise specified, measurements were performed manually by a trained, blinded researcher.

Open Field Test/NPR Task-

From day 1 to day 2, mice were placed into an empty open field chamber (45 cm × 45 cm × 45 cm constructed of grey PVC) for 10 min, referred to as the habituation phase of the NPR Task. The first 5 minutes of day 1 habituation were recorded with a video tracking system for the Open Field Test. EthoVision XT (Noldus, Netherlands) software was used to quantify locomotor activity (distance traveled) and time spent in the center of the open field (10cm from the nearest wall) as measures of anxiety-like behavior.

The encoding phase of the NPR was on day 3, during which mice were given 5 min to explore two identical objects in the chamber. 24h later was the retrieval phase, during which one of the two objects from the encoding phase was moved to a different location and mice once again had 5 min to explore. Objects for exploration were glass cylinders (height 4 cm, base diameter 1.5 cm) stuck to the arena floor in random arrangements between experiments at least 5 cm equidistant from the walls and at least 25 cm away from each other to ensure that corner preferences did not bias exploration times. For DREADD experiments, CNO at 1 mg/kg was i.p. injected 1.5 hr before the encoding phase, and for refeeding experiments, food was reintroduced 1h. before then encoding.

Object exploration was considered whenever the mouse sniffed the object or touched the object while looking at it (when the distance between the nose and the object was less than 1 cm). Climbing onto the object did not qualify as exploration. Exploration times were converted into a discrimination ratio (time at novel − time at old)/(time at old + time at novel), wherein a value of zero indicates no exploration preference, and a positive value indicates preferential exploration of the novel configuration, thus indicating memory of the familiar configuration. Any mice that showed preference for specific objects in the encoding phase (discrimination ratio ≥ 0.33), or did not explore both objects at least twice during either phase, were excluded from the analysis.

Zero Maze-

At least 2h. after NPR retrieval, mice were individually placed on an elevated white plastic ring platform with both open and closed arms (width of ring 6 cm, outer diameter 60 cm, 60cm off the ground) for 5min each. Mice were originally placed in the entryway of a closed arm and time spent in the open arm sections was scored as a measure of anxiety-like behavior.

Contextual Fear Conditioning-

On days 5–6 contextual fear conditioning experiments were carried out in a commercial fear conditioning system (Med Associate) comprised of a 29 × 25 × 22 cm chamber with grid floors, opaque ceilings, and white lighting. On Day 5, mice were placed in the chamber for 320s, during which they experienced two foot shocks (0.65 mA, 2 s) delivered at 180 s and 240 s. Memory retrieval was performed 24 hours later by placing the mice in the same chamber for 5 min and VideoFreeze software was used to quantify “freezing.”

Forced Swim-

On day 7, mice were placed in an acrylic cylinder (diameter 20 cm, height 30 cm) filled with room temperature water to a depth of 20 cm for 5 min. Time spent immobile, defined as all four paws being immobile, was measured as a proxy of depressive-like behavior.

Fiber photometry recording

Photometry recordings were performed as described previously²³ and fiber placement was verified in each animal. Briefly, after 10min acclimation, recordings were carried out in an open-top home cage (21.6 “ 17.8 “ 12.7 cm) in the 30 Lux red light environment using a multichannel spectrometer-based system. Photometry data was collected at 10 data points/s. and exported to MATLAB R2022b for analysis. Fluorescence bleaching and movement artifacts were corrected for by normalizing to TdTomato control and 10s. moving minimum windows (Figure S1P–S1R). Photometry signals (△F/F) were derived by calculating (F–F0)/F0, where F0 is the median of the fluorescence signal. Fluorescence “spikes” were defined as 3 standard deviations above the mean and AUC was calculated from the product of spike number and mean spike amplitude.

For perievent calcium analysis during NPR, photometry recordings were performed during standard NPR testing. “Events” were defined as isolated events during which mice investigated objects and, to prevent confounding effects from previous or subsequent object investigations, only isolated investigations with no other investigations within 7s. were analyzed. Perievent calcium analysis included peak fluorescence and AUC (mean fluorescence × time), normalized to baseline fluorescence (4s. before investigation).

Code for spectral unmixing of fluorescence signals and spike analysis can be found at: https://doi.org/10.5281/zenodo.16703985.⁶¹

2-Photon Microscopy in Acute Brain Slice

Slice Preparation-

Slice preparation was performed as described previously^64,65. Briefly, mice were anesthetized with 5% isoflurane in oxygen and transcardially perfused with ice-cold NMDG-based aCSF (N-methyl-D-glucamine) saturated with 95% O2 and 5% CO2 and containing (in mM): 92 NMDG, 30 NaHCO3, 25 glucose, 20 HEPES, 10 MgSO4, 5 sodium ascorbate, 3 sodium pyruvate, 2.5 KCl, 2 thiourea, 1.25 NaH2PO4, 0.5 CaCl2 (pH 7.3, 305–315 mOsm). Brains were rapidly removed, and acute coronal slices (300 μm) containing the hippocampus were cut using a vibratome (VT1200, Leica, Germany). Slices were warmed to 34.5°C for 7 min and then maintained in a holding chamber containing HEPES aCSF (in mM): 92 NaCl, 30 NaHCO3, 25 glucose, 20 HEPES, 5 sodium ascorbate, 3 sodium pyruvate, 2.5 KCl, 2 thiourea, 2 MgSO4, 2 CaCl2, 1.25 NaH2PO4 (pH 7.3, 305–315 mOsm) at room temperature for at least 1 hr before recording in artificial CSF (aCSF) at 22–24°C. For interneurons, aCSF contained (in mM): 125 NaCl, 26 NaHCO3, 10 glucose, 2.5 KCl, 2 CaCl2, 1.3 MgCl2, 1.25 NaH2PO4, (pH 7.24, 300 mOsm, bubbled with 95% O2 and 5% CO2). For GC’s CaCl2 and MgCl2 concentrations were adjusted to 2.2 and 1.1mM, respectively. Slices acclimated to aCSF for >10min prior to recordings.

Imaging and Analysis-

GCAMP recordings were performed and analyzed as described previously⁶⁵. Briefly, images were collected at <2s/cycle using an Olympus FVMPE-RS Multiphoton system. GCAMP ROI’s were identified and quantified using ImageJ and then exported to MATLAB R2022b for analysis. Fluorescence bleaching and movement artifacts were corrected for by normalizing to non-GCAMP-expressing control ROI’s in each slice and 10s. moving minimum windows. GCAMP signals (△F/F) were derived by calculating (F–F0)/F0, where F0 is the median of the fluorescence signal. Fluorescence “spikes” were defined as 2 standard deviations above the mean and AUC was calculated from the product of spike number and mean spike amplitude.

Drug Treatments

ICV Shikonin was administered as previously described⁴⁶. Briefly, 2.5 uL shikonin (0.5mM; Selleck) or vehicle (5% DMSO) was delivered via Hamilton syringe either daily in stHFD mice or every other day in DIO mice. Shikonin was prepared immediately before injections.

Brain Immunohistology

Animals were transcardially perfused with ice cold PBS followed by 4% PFA before brain removal. Brains incubated in 4% PFA at 4°C for 24h before at least 2 days in 30% sucrose. Brains were sliced in 40um serial sections using a sliding microtome to collect hippocampal brain regions and stored in anti-freeze solution at −20°C. Immunohistology was performed as previously described^23,66. Briefly, serial sections were permeabilizated with 0.5% Triton-100 TBS for 30min, before two 10 min washes with 0.05% Triton-100 TBS (TBST), followed by blocking with 3.33% fresh donkey serum in TBST for 1h., slices were incubated for 1–2 days, shaking, at 4°C in TBST containing primary antibodies. Primary antibodies against c-Fos (1:3000; Anti-Guinea Pig; Synaptic Systems), pro-CCK (1:1000; Anti-Rabbit, Frontier), PKM1 (1:500; Anti-Rabbit; Cell Signaling), GLUT1 (1:500; Anti-Rabbit; Cell Signaling), pH^t11 (1:500; Anti-Rabbit; Abcam), PKM2 (1:500; Anti-Rabbit; ThermoFisher), and pPKM2^ser37 (1:500; Anti-Rabbit; Signalway) were used. Sections were then washed twice for 5min. in TBST and then incubated with fluorescent labelled secondary antibodies (1:500; Invitrogen) at room temperature for 2 hr. Finally, slices were washed twice for 10 min. in TBST, incubated in DAPI staining solution (ThermoFisher, 1:5000 in PBS) for 20min, and mounted on microscope slides.

Imaging acquisition and analysis

Images were obtained using an Olympus FV3000 microscope using a 10x objective or 20x objective 2x zoom. Imaging analysis was performed using Fiji/ImageJ unless otherwise specified. For pPKM2^ser37 in GC’s, Imaris image analysis software was used. ROIs were manually drawn using hippocampal landmarks to distinguish between different regions so quantifications could be normalized to volume. Hippocampal sections spanned from the dorsal DG to the start of the ventral DG. For immunohistological staining experiments 5 hippocampal sections per animal were quantified. For AAV-mediated cell labeling, only sections showing viral expression were analyzed.

Chemogenetic manipulation

For acute chemogenetic experiments, a single I.P. injection of CNO at 1mg/kg was performed 1.5h. before NPR testing. For chronic chemogenetic experiments, we administered CNO via drinking water with 1% sucrose at ~2.0mg/kg (assuming mice drink ~2.0mL/day) for a total of 10 weeks. Fresh, light-protected water bottles containing CNO-infused drinking water were replaced every 3 days.

snRNA-seq analysis from published datasets

Single nuclei counts, cell metadata, and library metadata from Langlieb, et. al, 2023⁶² were downloaded from https://docs.braincelldata.org/downloads. Libraries originating from dentate gyrus dissectates were subsetted from the whole brain anndata object and used to create a Seurat v5 object. Scaling, normalization, dimension reduction, and clustering was performed using the Seurat SCTransform workflow⁶⁷. Published cell classes from Langlieb, et. al, 2023 were used to identify the bulk of cell populations. Identities of hippocampal-specific neuron populations not labeled by the original authors were identified using markers from Cembrowski, et. al, 2016⁶⁸. Immature granule cell neuroblasts were identified by the expression of Dcx. The cell clusters expressing markers of CA principal neurons were excluded to limit analysis to dentate gyrus-specific cell populations.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistics were performed using GraphPad PRISM 9 and statistical details of experiments are reported in figure legends (including statistical tests used, exact value of n, what n represents, and exact value of p). Individual animals are treated as biological replicates unless otherwise noted in figure legends. Data are all presented as mean +/− S.E.M. by animals unless otherwise noted in figure legends. No data were excluded unless specifically specified. Both behavioral analysis and cell counting were performed blinded to the conditions of the experiments. To compare the cell density, discrimination ratios, freezing percent, and other behavioral tests in different groups, we used unpaired t-tests or One-Way ANOVA’s as appropriate. Differences in body weight or GCAMP over time were determined using Two-Way ANOVA’s. Tukey’s, Dunnett’s, or Bonferroni’s corrections for multiple comparisons were used when appropriate. A two-tailed p-value < 0.05 was considered statistically significant.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-guinea pig c-Fos	Synaptic Systems	RRID: AB_2800522
Anti-rabbit PKM2	ThermoFisher	RRID: AB_2546176
Anti-rabbit pPKM2ser37	Signalway	Cat#: 11456-2
Anti-rabbit PKM1	Cell Signaling	RRID: AB_2715534
Anti-rabbit pHt11	Abcam	RRID: AB_304759
Anti-rabbit pro-CCK	Frontier	RRID: AB_2571674
Anti-rabbit GLUT1	Cell Signaling	RRID: AB_3064908
Alexa Fluor 647 Donkey anti-Rabbit	Invitrogen	RRID: AB_2762835
Alexa Fluor 568 Donkey anti-Rabbit	Invitrogen	RRID: AB_2534017
Alexa Fluor 488 Donkey anti-Rabbit	Invitrogen	RRID: AB_2535792
Alexa Fluor 647 Goat anti-Guinea pig	ThermoFisher	RRID: AB_2535867
Alexa Fluor 488 Donkey anti-Goat	Jackson Immuno	RRID: AB_2340472
Bacterial and virus strains
pENN-AVV5-CAMKII-GCAMP6f	James M Wilson (unpublished)	Addgene viral prep # 100834-AAV5
AAV9-hsyn-FLEX-jGCAMP7f	Dana et al. 2019 69	Addgene viral prep # 104492-AAV9
AAV5-CAG-tdTomato	UNC Vector Core/Boyden Lab	N/A
AAV2-hsyn-DIO-hm3dq-mcherry	Krashes et al. 70	Addgene viral prep # 44361-AAV2
AAV2-hsyn-DIO-hm4di-mcherry	Krashes et al. 70	Addgene viral prep # 44362-AAV2
AAV2-hsyn-DIO-mcherry	Bryan Roth (unpublished)	Addgene viral prep # 50459-AAV2
pAAV2[miR30]-SYN1>mCherry:{mPkm[miR30-shRNA]}:WPRE	VectorBuilder	shRNA targeting sequence for Pkm2: GATGTCGACCTTCGTGTAAAC46
pAAV2[miR30]-SYN1>mCherry:Scramble[miR30-shRNA]}:WPRE	VectorBuilder	Cat#: VB231005
Biological samples
N/A
Chemicals, peptides, and recombinant proteins
CNO	NIH	Cat#: C-929
Shikonin	SelleckChem	Cat#: S8279
Critical commercial assays
N/A
Deposited data
Original data reported in this paper will be shared by the lead contact upon request
Experimental models: Cell lines
N/A
Experimental models: Organisms/strains
VGAT-ires-Cre	Jackson laboratory	RRID: IMSR_JAX:016962
CCK-ires-Cre	Jackson laboratory	RRID: IMSR_JAX:012706
NPY-ires-Cre	Jackson laboratory	RRID: IMSR_JAX:027851
Pvalb-ires-Cre	Jackson laboratory	RRID: IMSR_JAX:017320
Oligonucleotides
shRNA targeting sequence for Pkm2: GATGTCGACCTTCGTGTAAAC	Pan et. al. 202246	N/A
Recombinant DNA
N/A
Software and algorithms
MATLAB R2022b	Mathworks	https://www.mathworks.com/help/install/ug/install-products-with-internet-connection.html
Imaris		https://imaris.oxinst.com/imaris-viewer
FIJI	ImageJ	https://imagej.net/software/fiji/downloads
Olympus FluoView	Evident Scientific	Evident Scientific
Prism10	Graphpad	https://www.graphpad.com/
Adobe Illustrator	Adobe	www.adobe.com
Original code for photometry analysis		https://doi.org/10.5281/zenodo.16703985
EthoVision XT	Noldus	https://my.noldus.com/download/latest/ethovision-xt
Other
N/A

Highlights.

• Short-term high-fat diet (stHDF) impairs memory via dentate CCK-IN hyperactivity

• Dentate CCK-INs are glucose inhibited and become hyperactive in response to stHFD

• Glucose or PKM2 modulation restores dentate CCK-IN activity and memory performance

• Early PKM2 or CCK-IN intervention prevents obesity-related memory deficits