Endogenous formaldehyde is a memory-related molecule in mice and humans

Abstract

Gaseous formaldehyde is an organic small molecule formed in the early stages of earth's evolution. Although toxic in high concentrations, formaldehyde plays an important role in cellular metabolism and, unexpectedly, is found even in the healthy brain. However, its pathophysiological functions in the brain are unknown. Here, we report that under physiological conditions, spatial learning activity elicits rapid formaldehyde generation from mitochondrial sarcosine dehydrogenase (SARDH). We find that elevated formaldehyde levels facilitate spatial memory formation by enhancing N-methyl-D-aspartate (NMDA) currents, but that high formaldehyde concentrations gradually inactivate the NMDA receptor by cross-linking NR1 subunits to NR2B via the C232 residue. We also report that in mice with aldehyde dehydrogenase-2 (ALDH2) knockout, formaldehyde accumulation due to hypofunctional ALDH2 impairs memory, consistent with observations of Alzheimer's disease patients. We also find that formaldehyde deficiency caused by mutation of the mitochondrial SARDH gene in children with sarcosinemia or in mice with Sardh deletion leads to cognitive deficits. Hence, we conclude that endogenous formaldehyde regulates learning and memory via the NMDA receptor.

Keywords: Alzheimer's disease; Energy metabolism; Hippocampus.

Fig. 1. Spatial learning elicits formaldehyde generation.

a, b Spatial learning and memory in wild-type SD rats trained in the MWM (n = 10 per group). c Brain formaldehyde fluorescence revealed by the in vivo imaging system. NaFA: a fluorescence probe of free formaldehyde, n = 3. d Hippocampal formaldehyde (FA) levels detected by Fluo-HPLC (n = 10). e An in vivo LTP recording in CA1 from Schaffer collateral stimulation and 3D views of the hippocampus (yellow). SE stimulating electrode, RE recording electrode. f Late-LTP (L-LTP) formation in vivo. n = 8, HFS high-frequency stimulation. g Brain formaldehyde revealed by the in vivo imaging system, n = 3. h Hippocampal formaldehyde levels detected by Fluo-HPLC (n = 6). i The pathway of formaldehyde metabolism. SARDH sarcosine dehydrogenase, ALDH2 aldehyde dehydrogenase. j Colocalization of SARDH (red) and the mitochondrial marker- Cox IV (green) in the cultured hippocampal neurons. DAPI: blue, a nuclear dye. k, l The cultured medium sarcosine and formaldehyde levels detected by Fluo-HPLC. SA sarcosine, MA methoxyacetic acid, an inhibitor of SARDH, n = 6. m Intracellular infusion of the mitochondrial formaldehyde probe (mito-FA-probe). n A train of 20 pulses in the presynaptic neurons induced multiple action potentials in the postsynaptic neurons, n = 8. o The active formaldehyde generated from mitochondria in the axon, synapse and soma of the cultured neurons imaged by the mito-FA probe, n = 6. p The model of endogenous formaldehyde-enhanced memory formation. The data are expressed as the mean ± standard error (s.e.m.).  ***p < 0.001; ****p < 0.0001.

Fig. 2. Formaldehyde regulates LTP and memory via NMDA-R.

a The metabolism pathway of formaldehyde precursors. GAMT guanidinoacetate methyltransferase, SARDH sarcosine dehydrogenase (a formaldehyde-generating enzyme), ALDH2 aldehyde dehydrogenase (a formaldehyde-degrading enzyme), FA formaldehyde, CT creatine, SA sarcosine. b Hippocampal formaldehyde levels quantified by HPLC-Fluo 30 min after i.c.v. injection with medicines—creatine (200 μM), sarcosine (200 μM), and formaldehyde (50 μM) (n = 6). c, d The fEPSP traces at 60 and 120 min (c) and enhancement of hippocampal long-term potentiation (LTP) (d) by infusion of above medicines (n = 8). e After 6 days of MWM training, repeated measures two-way ANOVA revealed a difference in group: F_{(3, 36)} = 32.43, p < 0.001, time: F_{(5, 203)} = 74.98, p < 0.001, and a group/time interaction: F_{(15, 203)} = 6.32, p < 0.001. Post hoc tests showed that the mean escape latency values for the CT- and SA-injected group were significantly shorter than the control group on days 4, 5 and 6, t (36) = 5.17, p < 0.01; and FA-injected group had shorter escape latency than the control group on days 1, 2, 3, 4, 5 and 6, t (36) = 7.11, p < 0.01, respectively. f The CT-, SA-, and FA-injected mice had longer staying times in target quadrant than control (wild-type mice) (t (36) = 9.29, p < 0.001, n = 10 mice per group). g, h The effects of Ifen and APV on spatial memory assayed by MWM (n = 10, rat per group). Ifenprodil (Ifen, 10 μM, a specific antagonist of NR2B); DL-2-amino-5-phosphonovaleric acid (APV, 50 μM, a nonspecific antagonist of NMDA-R). i–l The effects of extracellular or intracellular infusion of various formaldehyde concentrations on NMDA currents in the cultured hippocampal neurons (n = 9−13). Data are expressed as the mean ± standard error (s.e.m.).  ***p < 0.001; ****p < 0.0001.

Fig. 3. Formaldehyde dually regulates NMDA-R via different subunits.

a, b The model of active formaldehyde-activated NMDA-R by binding to C232 residue of NR2B. Ifenprodil (Ifen, a specific antagonist of NR2B). FA formaldehyde. c The chemical equation between formaldehyde and l-cysteine. d, e Recording of NMDA currents in CHO cells cotransfected with GFP-NR1a and NR2B without the amino terminal domain (NR2B-ATD) or a single point mutation (C232A). LBD ligand binding domain. n = 10−12. f, g The model of excessive formaldehyde-suppressed NMDA-R by cross-linking C79 of NR1 to K79 of NR2B. h, i Recording of NMDA currents in CHO cells cotransfected with C79S mutant NR1a and/or K79S mutant NR2B, respectively (n = 10−12). The data are expressed as the mean ± standard error (s.e.m.). ***p < 0.001; ****p < 0.0001.

Fig. 4. ALDH2 mutation-induced formaldehyde overload causes amnesia.

a The scheme for generation of Aldh2^−/− mice. b After 6 days of MWM training, repeated measures two-way ANOVA revealed a difference in group: (F_{(2, 27)} = 14.19, p < 0.001), training time (F_{(5, 152)} = 58.32, p < 0.001), and a group × time interaction (F_{(10, 152)} = 4.91, p < 0.001). Post hoc tests showed that the mean escape latency values for ALDH2^−/− mice were a significant longer than control (wild-type mice) on day 3 (F_{(2, 27)} = 3.86, p = 0.005), day 4 (F_{(2, 27)} = 4.61, p = 0.004), day 5 (F_{(2, 27)} = 7.13, p = 0.002), and day 6 (F_{(2, 27)} = 6.27, p = 0.001); while the escape latency in ALDH2^−/− mice with l-cys treatment was no statistically significant difference than control from days 1 to 5 (p > 0.05; n = 10 mice per group). c ALDH2^−/− mice with l-cys injection reversed the reduced time in target quadrant of these mutated mice without formaldehyde treatment (t (27) = 6.25, p < 0.001). d Hippocampal formaldehyde (FA) concentrations detected by Fluo-HPLC (n = 10). e, f Spatial learning and memory assayed by the MWM in WT rats with or without intrahippocampal infusion of excessive (450 μM; t (27) = 11.60, p = 0.002) formaldehyde or formaldehyde scavenger-l-cysteine (l-cys: 500 μM) (t (27) = 1.49, p = 0.165); n = 10 mice/group). g Hippocampal formaldehyde levels detected by Fluo-HPLC (n = 10). h Negative relationship between urine formaldehyde and MMSE scores in 158 elderly AD patients. i ALDH2 activity analyzed by a human ALDH2 kit (p < 0.01). j Prefrontal lobe atrophy revealed by MRI. k Ventriculomegaly in AD patients with ALDH2 mutation. The data are expressed as the mean ± standard error (s.e.m.).  ***p < 0.001; ****p < 0.0001.

Fig. 5. SARDH mutation-induced formaldehyde deficiency causes amnesia.

a Positive relationship between urine formaldehyde (FA) levels and Wechsler Intelligence Scale for Children (WISC) scores in 11 pediatric sarcosinemia patients and 31 healthy children. b Blood SARDH activity analyzed by a human SARDH kit (p < 0.01). c Urine formaldehyde concentrations detected by Fluo-HPLC (p < 0.01). d Right hippocampal atrophy revealed by MRI. e Impaired glucose metabolism in the right hippocampus revealed by ¹⁸F-FDG PET/CT. f The scheme for generation of Sardh^−/− mice. g After 6 days of MWM training, repeated measures two-way ANOVA revealed a difference in group: (F_{(2, 27)} = 16.38, p < 0.001), training time (F_{(5, 152)} = 72.54, p < 0.001), and a group × time interaction (F_{(10, 152)} = 2.45, p < 0.001). Post hoc tests showed that the mean escape latency values for SARDH^−/− mice were a significant longer than control (wild-type mice) on day 3 (F_{(2, 27)} = 2.41, p = 0.001), day 4 (F_{(2, 27)} = 4.57, p = 0.005), day 5 (F_{(2, 27)} = 6.39, p = 0.003), and day 6 (F_{(2, 27)} = 6.43, p = 0.002); while there was no statistically significant difference in escape latency between SARDH^−/− mice with 200 μM formaldehyde injection and control (p > 0.05; n = 10 mice per group). h SARDH^−/− mice with formaldehyde infusion reversed the reduced time in target quadrant of these mice without formaldehyde treatment (p < 0.001). i Hippocampal sarcosine levels detected by a mouse sarcosine kit, n = 6. j Hippocampal formaldehyde levels detected by Fluo-HPLC, n = 6. k, l NMDA currents recording in the brain slices of Sardh^−/− mice and formaldehyde-injected Sardh^−/− mice, n = 10−12. m The effects of formaldehyde scavengers- l-cys (300 μM) and NaHSO₃ (300 μM) on the cell viability of the cultured hippocampal neurons detected by CCK-8 kit, n = 6. Data are expressed as the mean ± standard error (s.e.m.). ***p < 0.001; ****p < 0.0001.