All procedures involving the use of animals were approved by the Oregon Health & Science University (OHSU) Institutional Animal Care and Use Committee (IACUC) and in compliance with the U.S. Department of Agriculture Animal Welfare Act and the National Institutes of Health policy on Humane Care and Use of Laboratory Animals.

Thirty-six 4-month-old male mice of one of three genotypes: homozygous GPR39 KO, heterozygous (Het), or WT littermates were used in this study. The GPR39 mutant mice were generated by Cyagen Biosciences (Santa Clara, CA, United States) using a CRISPR/Cas9 deletion strategy and zygotic injections, which have previously been reported (Alkayed et al., 2022).

The animals were group-housed in an animal room colony (5 per cage) under a 12/12-h light/dark cycle (light onset at 06:00 = zeitgeber time [ZT] 0) at 21–23°C ambient temperature and 30–42% relative humidity. An overview of the experimental timeline is presented in Figure 1A. At 4 months of age, the animals were randomly assigned to either high-fat diet (HFD: 60 kcal% fat; D12492i, Research Diets, New Brunswick, NJ, United States) or standard (STD) chow (13 kcal% fat, PicoLab Laboratory Rodent Diet, 5LOD, LabDiet, St. Louis, MO, United States), with food and water available ad libitum throughout the study. After 8 months on the diets, cognitive performance was evaluated using the Novel Object Recognition (NOR) and the Morris water maze tests, followed by CBF measurement with and without hypercapnia, to assess vasoreactivity, using ASL MRI. Upon completion of the study, blood was collected by cardiac puncture, and mice were transcardially perfused with heparinized cold saline (1 unit/mL) to remove blood from circulation. Brain tissues were then removed and flash frozen for oxylipins analysis. Collected plasma samples were stored at −80°C for subsequent measurements of insulin and cytokines/chemokines.

Behavioral testing and MRI were completed by experimenters blinded to the experimental and treatment groups. Due to differences in food color and mouse appearance, HFD status could not be blinded during behavioral testing or MRI. All data collected were analyzed by investigators blinded to both diet and treatment status. Reporting of results conforms to the ARRIVE 2.0 (Animal Research Reporting in In Vivo Experiments) guidelines.

Mice were behaviorally tested over 14 days. Mice were tested for exploratory activity and measures of anxiety in the elevated zero maze on day 1 and the open field on days 2 and 3 (Supplementary Materials). Mice were then tested for novel object recognition on days 4 and 5, and for spatial learning and memory in the water maze on days 8 through 12. Finally, mice were tested for contextual and cued fear conditioning on days 13 and 14 (Supplementary Materials). Novel object recognition and water maze tests were performed as described below in detail.

Magnetic resonance imaging data acquisition and analysis: Imaging was conducted following behavioral testing. MRI was performed at the OHSU Advanced Imaging Research Center using a Bruker-Biospin 11.75 T small animal MR system with a ParaVision 5.1 software platform, 10 cm inner diameter gradient set with a 72 mm (ID) and 60 mm (length) RF resonator for transmitting and an actively decoupled mouse head surface coil for receiving. Mice were anesthetized with a ketamine/xylazine mixture (1.0 mg xylazine/7 mg ketamine/100 g) in combination with low isoflurane (0.75%) in 100% oxygen. The mice were positioned with heads immobilized on an animal cradle. Body temperature of the mice was monitored and maintained at 37°C while monitoring respiration (SA Instruments, Stony Brook, NY, United States). For each mouse, a coronal 25-slice T2-weighted image was acquired (ParaVision spin echo RARE, 256 × 256 matrix, 125 μm in-plane resolution, 0.5 mm slice width, TR 4000 ms, TE effective 23.64 ms, RARE factor 8, 2 averages). These T2-weighted anatomical scans were used for positioning the perfusion image slice at a consistent position approximately 1.75 mm anterior to the anterior commissure. Cerebral blood flow (ml/min/100 g) was measured using ASL, employing the flow-sensitive alternating inversion recovery rapid acquisition with relaxation enhancement pulse sequence (ParaVision FAIR-RARE), with TE/TR = 45.2/10000 ms, slice thickness = 2 mm, number of slices = 1, matrix = 128 × 128, 250 μm in-plane resolution RARE factor = 72, and 23 turbo inversion recovery values ranging from 40 to 4400 ms, and acquisition time of 15 min. This sequence labels the inflowing blood by global inversion of the equilibrium magnetization (Kim, 1995). The ASL sequence was implemented first at the baseline condition with 100% oxygen, and subsequently 10 min after switching to the hypercapnia condition with 95/5% oxygen/carbon dioxide to assess vasoreactivity (or cerebrovascular reactivity), which is a hemodynamic parameter representing the increase in normal cerebral artery blood flow in response to a vasodilatory stimulus such as hypercapnia. Resulting images were analyzed using the Bruker ParaVision software and JIM software (Xinapse Systems LTD, Northants, United Kingdom). Cerebral blood flow maps (ml/100 g-min) were generated using the Bruker ASL perfusion processing macro and exported into JIM for further processing. Outlier value brain pixels (outside 2 SDs) representing large arteries with high, pulsatile flow were excluded, thus arriving at flows which consistently represent tissue microvascular flow. The identical FOV geometry offsets of the T2 and ASL images enabled the ROIs drawn on the T2 image to be readily overlaid onto the corresponding perfusion map for quantification. Mean cerebral blood flow was quantified for the whole brain and for each brain hemisphere, each defined anatomically using the corresponding T2 images.

Following the acquisition of the MRI data, blood was collected through cardiac puncture at the time of euthanasia. Blood was centrifuged at 2,000 rpm for 15 min to separate the plasma, which was stored at −80°C until assay.

Unless indicated otherwise, data were analyzed by analysis of variance (ANOVA) with Tukey’s and Dunnett’s multiple comparison post-hoc tests using diet as the between-subject factor. G*Power 3 (La Frano et al., 2017) was used to calculate the power analysis. As we a priori expected a genotype-dependent response to diet, we analyzed the effect of diet in each genotype separately. The number of mice needed to detect an effect of diet for each of the three genotypes based on our previous studies (Johnson et al., 2016; Zuloaga et al., 2016; Davis et al., 2021b) was 12 mice per genotype per diet. All data are shown as mean ± standard error of the mean (SEM) and were first assessed for normality of variance. Statistical analyses were performed using SPSS v.27.0 (IBM Corp., Armonk, NY, United States) and Prism v.9.2.0 (GraphPad Software Inc., San Diego, CA, United States) software. If a violation of sphericity occurred indicating that the variances of the differences between all combinations of the groups were not equal, Greenhouse-Geisser corrections were used. All figures were generated using GraphPad Prism software and we considered p < 0.05 as statistically significant.

Linear regression analysis was used to compare the treatment groups in terms of oxylipins values investigating for which oxylipins there was significant difference between treatments. Log transformation of outcome values was performed to improve the model fit in terms of model assumptions and few outliers were excluded from statistical analyses. For Pathway data, for each class of fatty acids, the oxylipins values were summed up based on common pathway first and then linear regression model was fitted to compare the treatment groups for each pathway. Post-hoc analysis was performed to compare some groups of interest (e.g., WT vs. GPR39 KO mice) using the Benjamini-Hochberg false discovery rate correction method for multiple comparisons t-tests. The results (coefficients, estimates, standard errors, and corresponding p-values, and fold changes) are provided in Tables 1–4. Analyses were conducted using R 4.1.0 (R core Team, Vienna, Austria) (R Core Team, 2021).

High-fat diet caused an increase in body weight in both genotypes, resulting in significant weight gains compared to mice on STD diet at the end of the study (WT: HFD 56.20 ± 1.39 vs. STD 37.25 ± 1.93; KO: HFD 56.00 ± 0.45 vs. STD 35.80 ± 1.24; Figure 1B). No differences were observed in starting weights or weight gain on HFD between WT and KO mice. Fasting blood glucose was significantly higher in mice on HFD vs. STD in both genotypes, with no differences between the two genotypes (Figure 1C). However, plasma insulin levels were significantly reduced in KO mice regardless of diet (Figure 1D). The latter finding is consistent with previous reports indicating that GPR39 is required for the increase insulin secretion under conditions of increased demand, such as age-dependent and diet-induced insulin resistance (Tremblay et al., 2009). The discrepancy between the changes in glucose and insulin in GPR39 KO mice has previously been observed by others (Tremblay et al., 2009). GPR39 augments insulin secretion under conditions of increased demand; e.g., age-dependent and diet-induced insulin resistance. The effect is in part related to the role of GPR39 in pancreatic cell hyperplasia associated with the development of insulin resistance. The contribution of GPR39 to insulin secretion and plasma glucose regulation is revealed under conditions of high demand and reduced β-cell function, where the increase in insulin resulting from GPR39 deletion would be matched by a corresponding rise in glucose.

When activity levels were assessed in the open areas of the elevated zero maze, there was no effect of HFD in any genotype. However, HFD reduced activity in the closed areas of the elevated zero maze in Het mice (Supplementary Figure 1A). This effect was not seen in WT or KO mice. In contrast to activity levels, there was no effect of HFD on measures of anxiety in the elevated zero maze in any genotype. Consistent with the elevated zero maze data, there was a trend for a reduction in activity levels in the open field in Het mice on a HFD (Supplementary Figure 1B), with no effects on activity levels seen in WT or KO mice. Finally, there was a trend for a decrease in Het mice for time spent in the anxiety-provoking center of the open field (Supplementary Figure 1C) and the number of entries into the center (Supplementary Figure 1D).

The time spent exploring the novel and familiar object in NOR was analyzed. There was a trend for KO mice on a HFD to spend more time exploring the novel object, but the difference did not reach statistical significance (p = 0.06969, Supplementary Figure 2). In the visible platform training of the water maze test, swim speed was lower in Het mice on HFD (Figure 2A), with a similar trend in KO mice, but the decrease did not reach statistical significance (p = 0.071, Figure 2A). There was no effect of diet on latency or cumulative distance to target in any genotype during visible platform training (Supplementary Figure 3). During hidden platform training, cumulative distance to target was significantly higher in KO mice compared to WT or Het, regardless of diet (Figure 2B). Because of the effect of HFD on swim speeds, we also analyzed distance moved as a performance measure. HFD did not reduce the ability to locate the visible platform in any genotype. During the hidden platform training, there was a trend for HFD to reduce distance moved in WT mice that did not reach statistical significance (p = 0.081). Next, spatial memory retention was assessed in the water maze probe trial (no platform). In WT mice on STD there was an effect of quadrant (p = 0.092), with a trend for more time spent in the target quadrant that was not statistically significant (Figure 2C). Interestingly, a strong spatial bias was seen in WT on the HFD, with mice spending significantly more time in the target quadrant than any other quadrant (p = 0.003). Het mice on either STD or HFD also showed spatial bias and spent more time in the target quadrant than any other quadrant (Figure 2C). In contrast, no effect of quadrant was seen in KO mice regardless of diet; the mice spent a similar amount of time in all quadrants (Figure 2C). Finally, fear learning and memory were assessed. When activity levels were analyzed during the baseline period (prior to the first tone), there was an effect of diet in Het mice, with lower activity levels in Het mice on a HFD in fear conditioning (Supplementary Figure 4A). This effect was genotype-dependent and not seen in WT or KO mice. When motion during the shocks was analyzed, there was an effect of diet in WT mice, with lower responses to the shock in WT mice on a HFD (Supplementary Figure 4B). This was not seen in Het or KO mice. There was no effect of diet on freezing during the tones or the inter-stimulus intervals in any genotype. There was no effect of diet on freezing during the contextual or cued fear memory tests (Supplementary Figure 4C). Taken together, these results indicate that GPR39 deletion impairs spatial memory retention and suggest a protective role for GPR39 against cognitive impairment.

Because there was no observed cognitive deficit in the Het group, we decided to remove it from the rest of the study. To characterize the differences in inflammatory state in mice, 26 protein markers of systemic inflammation were measured in plasma (Supplementary Table 1). Cytokine/chemokines multiplex analysis showed a significant interaction between genotype and diet and a significant diet effect (Figure 3A). The post-hoc analysis revealed a significant difference in IL-10 between KO mice on HFD vs. STD, with no such difference in WT. There was a statically significant difference in IL-4 between mice on STD vs. HFD (Figure 3B), with no statistically significant differences between WT and KO. Similarly, IP-10 was significantly higher in mice on HFD vs. STD (Figure 3C), with differences between WT and KO. Lastly, MCP-3 was higher in WT on HFD vs. STD, with no diet effect in the KO (Figure 3D). Regardless of diet, a trend for a decrease in IL-27 (p = 0.0523) in KO was observed without reaching statistical significance.

We next investigated whether GPR39 deficiency altered levels of specific oxylipins in brain and plasma. The data spread for oxylipins levels in the brain (in nanomoles, nM) is presented as the interquartile range (IQR) in Supplementary Tables 2, 3. Linear regression analysis was performed to compare experimental groups. Brain and plasma oxylipins were grouped according to the main biosynthetic pathways by (1) fatty acid (FA) precursors (i.e., LA, ARA, EPA, and DHA), oxylipin groups (i.e., midchain hydroxy-octadecadienoic acid (HODE), epoxy-eicosatrienoic acid (EET), mid-chain hydroxy-eicosatetraenoic acid (HETE), epoxy-docosapentaenoic acid (EpDPA), dihydroxy-docosapentaenoic acid (DiHDPA), (2) enzymes involved in their synthesis [i.e., oxygenation of PUFAs by LOX followed by reduction, or alternatively hydroxylation of PUFAs by CYP1B1; oxidation of PUFAs by CYP450 followed by hydroxylation of oxidized PUFAs by soluble epoxide hydrolase (sEH)], and (3) based on enzymatic product-to-substrate ratio (i.e., hydroxylation of 10,11-EpDPA to 10,11-DiHDPA, 14,15-EET to 14,15-DiHET, or 19,20-EpDPA to 19,20-DiHDPA by sEH) and reported in Tables 1, 2. Brain oxylipins analysis showed decreases in the concentrations of 11(S)-HETE, 15(S)-HETE, PGE2, PGD2, PGJ2, 6-keto PGF1a, 8-iso PGF2a, 11 b-PGF2a, 15-keto, PGF2a, 13,14-dihydro-15-keto PGE2, 13,14-dihydro-15-keto PGD2, 15-deoxy-delta12,14-PGJ2, thromboxane B2, 14,15-DiHET, and 7,8-DiHDPA in KO mice fed with HFD compared to WT mice fed with HFD. However, brain oxylipins showed increases in the concentrations of 11(S)-HETE, 15(S)-HETE, PGE2, PGD2, PGJ2, 6-keto PGF1a, 8-iso PGF2a, 11 b-PGF2a, 15-keto, PGF2a, 13,14-dihydro-15-keto PGE2, 13,14-dihydro-15-keto PGD2, 15-deoxy-delta12,14-PGJ2, Thromboxane B2, 14,15-DiHET, and 7,8-DiHDPA when comparing WT mice fed with HFD to WT mice fed with STD. While only PGE2, PGD2, 15-keto PGF2a, and Resolvin D1 appeared to be increased in KO mice fed with STD compared to WT mice fed with STD (Table 1). Fifteen oxylipins in the AA pathway identified above decrease significantly in response to HFD and GPR39 deletion, while only Resolvin D1 and 7,8-DiHDPA in the DHA pathway were identified to increase in response to HFD and GPR39 KO for individual oxylipins. Except Resolvin (D1, D2, and D3) in the 12/15-LOX pathway which increased in KO mice fed with HFD compared to WT mice fed with HFD, six oxylipins 11(S)-HETE, 15(S)-HETE, PGE2-PGD2-PGJ2, 6-keto PGF1a, 8-iso PGF2a, and 15-deoxy-delta12,14-PGJ2 in the AA pathway were identified to significantly decrease in response to HFD in KO mice compared to WT mice, while increasing in WT mice fed with HFD compared to WT mice fed with STD. In addition, prostaglandins (E₂, D₂ and J₂) significantly increased in KO mice on STD compared to WT mice on STD (Table 2). The changes in the composition of these brain oxylipins were mainly associated with AA-derived oxylipins pathway. Like the brain, plasma oxylipins were also analyzed and results are presented in Tables 3, 4. Individual plasma oxylipins show a significant decrease in 5,6-DiHETE (EPA pathway) in KO mice on HFD compared to WT mice on HFD (Table 3). In addition, a significant effect of HFD was only seen in WT mice for 15 metabolites within five pathways: ARA, DHA, EPA, LA, and ALA (Table 3). Interestingly, HFD was associated with a significant increase in 20-HETE and a significant decrease in PGD₂ in WT mice. Levels of 20-HETE-11,12-EET-14,15-EET (CYPEPOX pathway) were reduced in KO mice on HFD compared to WT mice on HFD (Table 4). The 20-HETE-11,12-EET-14,15-EET (CYPEPOX pathway) was increased by HFD in WT mice, while four metabolites in the ARA, DHA, EPA, and LA pathways were decreased by HFD only in WT mice (Table 4). Taken together, these results highlight changes in brain and plasma oxylipins by HFD and GPR39 gene deletion that are consistent with inflammation.

At baseline, HFD reduced CBF in the entire brain (Figure 4A) with no difference between WT and KO. The effect of HFD persisted when CBF was analyzed for each hemisphere separately; again, without difference between WT and KO (Figures 4B,C). With hypercapnia (CO₂ challenge), the absolute mean CBF increased in all groups with no effect of HFD diet or GPR39 deletion for the whole brain CBF (Figure 4D) and for CBF in the left and right hemispheres (Figures 4E,F). When whole brain CBF was expressed as a percent change from baseline in response to hypercapnia [reflecting cerebrovascular reactivity (CVR) to CO₂ challenge], we observed higher CVR in HFD vs. STD in both genotypes, with no differences between WT and KO mice (Figure 5A). Similarly, when evaluated separately for left and right brain hemisphere, the augmenting effect of HFD but not genotype on CVR was still observed (Figures 5B,C). Overall, these results indicate that HFD is associated with a significant reduction in resting CBF, but paradoxically, a significant increase in the ability to dilate, likely reflecting a larger difference between the lower baseline and maximum dilation (Figure 5D).