Dietary salt promotes cognitive impairment through tau phosphorylation

Abstract

Dietary habits and vascular risk factors promote both Alzheimer's disease and cognitive impairment caused by vascular factors^1-3. Furthermore, accumulation of hyperphosphorylated tau, a microtubule-associated protein and a hallmark of Alzheimer's pathology⁴, is also linked to vascular cognitive impairment^5,6. In mice, a salt-rich diet leads to cognitive dysfunction associated with a nitric oxide deficit in cerebral endothelial cells and cerebral hypoperfusion⁷. Here we report that dietary salt induces hyperphosphorylation of tau followed by cognitive dysfunction in mice, and that these effects are prevented by restoring endothelial nitric oxide production. The nitric oxide deficiency reduces neuronal calpain nitrosylation and results in enzyme activation, which, in turn, leads to tau phosphorylation by activating cyclin-dependent kinase 5. Salt-induced cognitive impairment is not observed in tau-null mice or in mice treated with anti-tau antibodies, despite persistent cerebral hypoperfusion and neurovascular dysfunction. These findings identify a causal link between dietary salt, endothelial dysfunction and tau pathology, independent of haemodynamic insufficiency. Avoidance of excessive salt intake and maintenance of vascular health may help to stave off the vascular and neurodegenerative pathologies that underlie dementia in the elderly.

Extended Data Fig. 1:. HSD (4 and 8%) induced tau phosphorylation: brain localization, sex differences, and time course

a, HSD (NaCl 8%) increases tau phosphorylation on Ser³⁹⁶ (PHF13) and on Ser¹⁹⁹Ser²⁰² in the hippocampus but not in the neocortex (HIPP: PHF13, ND/HSD n=4/5, *p=0.0016 vs ND, Ser¹⁹⁹Ser²⁰², ND/HSD n=4/5, *p=0.0337, two-tailed unpaired t-test), whereas acetylation of Tau on Lys²⁸⁰ (K280) is not affected. MC1 levels increase in both neocortex and hippocampus but reaching statistical significance only in neocortex (CTX: MC1, ND/HSD n=4/5, *p=0.0321 vs ND, two-tailed unpaired t-test). b, Tau phosphorylation on Ser¹⁹⁹Ser²⁰² and Ser²⁰²Thr²⁰⁵ is abolished after treatment of brain samples with lambda phosphatase. c, HSD increases AT8 in neocortex and in hippocampus of female mice, whereas RZ3 only increases in neocortex (AT8, CTX: ND/HSD n=8/10, *p=0.0159 vs ND; HIPP: ND/HSD n=10/9, *p=0.0151 vs ND; RZ3, CTX: ND/HSD n=10/8, *p=0.0117 vs ND, two-tailed unpaired t-test). HSD induces a deficit in novel object recognition in female mice (ND/HSD n=8/9, *p=0.0017 vs ND, two-tailed unpaired t-test). d, HSD increases AT8 immunoreactivity in neuronal cell bodies of the somatosensory cortex (size bar=100 μm; 10μm in inset) and MC1 immunoreactivity in neuronal bodies of the pyriform cortex (size bar=500 μm; 100 μm in inset). Representative images from ND and HSD mice (n=5/group). e, Thioflavin S staining is not present in mice fed a HSD indicating absence of neurofibrillary tangles, which can be observed in rTg4510 mice (size bar=500μm; 100 μm in inset). Representative images from n=5/each ND and HSD mice and n=3 rTg4510 mice. f, HSD (4%) increases AT8 levels in neocortex but not in hippocampus (AT8, CTX: ND/HSD n=5/5, *p=0.0148 vs ND, two-tailed unpaired t-test). RZ3 was not increased in both regions. g, Time course of the increase in AT8 and RZ3 induced by HSD in the hippocampus. AT8 levels are increased at 4 and 12 weeks of HSD. RZ3 levels are increased at 4, 12, 24 and 36 weeks of HSD (AT8, 4 weeks: ND/HSD n=4/5, *p=0.0386 vs ND; 12 weeks: ND/HSD n=9/9, *p<0.0001 vs ND; RZ3, 4 weeks: ND/HSD n=4/5, *p=0.0041 vs ND; RZ3, 12 weeks: ND/HSD n=9/9, *p=0.0011 vs ND; 24 weeks: ND/HSD n=7/10, *p=0.0017 vs ND; 36 weeks: ND/HSD n=5/4, *p=0.0188 vs ND, two-tailed unpaired t-test). For gel source data see supplementary figure 1. Data are expressed as mean±SEM.

Extended Data Fig. 2:. Effect of HSD on neurons, astrocytes, microglia/macrophages, pericytes, and white matter integrity.

a, HSD (NaCl 8%) does not affect neurons (Neun), astrocytes (GFAP), microglia/macrophages (Iba-1, CTX: ND/HSD n=5/5, p=0.0570 vs ND; HIPP: ND/HSD n=5/5, p=0.0556 vs ND, two-tailed unpaired t-test) and pericytes (CD13) in both the pyriform cortex and the hippocampus (size bar=500 μm). b, No evidence of neuronal cell death is observed in HSD mice by using Fluoro-Jade B or TUNEL stainings (size bar=500 μm). +DNase indicates positive control for TUNEL staining. Representative images from ND and HSD mice (n=5/group). c, Kluver-Barrera stain shows no white matter damage in the corpus callosum of HSD mice. Representative images from ND and HSD mice (n=4/group). Data are expressed as mean±SEM.

Extended Data Fig. 3:. Aβ levels in HSD and correlation of behavioral deficits with p-tau, as well as p-tau in hypertension, HSD-treated tg2576 mice and hypothermia,

a,HSD (NaCl 8%) does not alter the distance travelled before finding the escape hole (Primary Distance, ND/HSD n=13/13, Diet: *p=0.0462, Time: *p<0.0001, two-way RM ANOVA plus Bonferroni’s test; Primary Distance Day 5: ND/HSD n=13/13, p=0.0670 vs ND, two-tailed unpaired t-test) or the number of errors made (Primary Errors, ND/HSD n=13/13, Diet: p=0.110, Time: *p=0.0004, two-way RM ANOVA plus Bonferroni’s test; Primary Errors, Day 5: p=0.1226 vs ND, two-tailed unpaired t-test). b, RZ3 levels in the cortex correlate with the cognitive performance at the novel object recognition test. No correlation was found between hippocampal RZ3 levels and cognitive performance at both the Barnes Maze and the novel object recognition test (RZ3 CTX: BM r=0.2828, *p=0.0133, n=76; NOR r=−0.2806, *p=0.0170, n=72; RZ3 HIPP: BM r=0.1739, p=0.1470, n=71; NOR r=−0.1746, p=0.1577, n=67, Pearson’s correlation coefficient). c: HSD did not increase soluble or insoluble Aβ₃₈, Aβ₄₀ or Aβ₄₂ in neocortex (Aβ₃₈, Soluble ND/HSD n=11/9, Insoluble ND/HSD n=7/6; Aβ₄₀, Soluble ND/HSD n=11/14, Insoluble ND/HSD n=7/6; Aβ₄₂, Soluble ND/HSD n=9/9, Insoluble ND/HSD n=7/6). d, Delivery of ANGII (600ng/kg·min, s.c.) with osmotic minipumps over 6 weeks increases systolic blood pressure and induces cognitive deficits (SBP – Veh/ANGII n=10/10, Treatment: *p<0.0001, Time: *p<0.0001, repeated two-way ANOVA and Bonferroni’s test; NOR – 2 weeks Veh/ANGII n=12/12, 4 weeks Veh/ANGII n=10/11, 6 weeks Veh/ANGII n=7/7, Treatment: *p<0.0021, Time: *p=0.0208, two-way ANOVA and Bonferroni’s test), e, ANGII administration increases AT8 and RZ3 in neocortex but not hippocampus (CTX, AT8 6 weeks: Veh/ANGII n=4/4, *p=0.0324 vs Veh; RZ3 6 weeks: Veh/ANGII n=5/5, *p=0.0262 vs Veh; HIPP, AT8 6 weeks: Veh/ANGII n=5/5, p=0.4056; HIPP, RZ3 6 weeks: Veh/ANGII n=5/5, p=0.0556, two-tailed unpaired t-test). f, HSD increases AT8 and RZ3 levels in both the neocortex and the hippocampus of 6 months-old Tg2576 mice (CTX, AT8: *p<0.0001 vs ND; HIPP, AT8: *p=0.0153 vs ND; CTX, RZ3: *p<0.0001 vs ND; HIPP, RZ3: *p=0.0239 vs ND, two-tailed unpaired t-test). g, Hypothermia induces massive AT8 phosphorylation (CTX: AT8 n=4/5, *p=0.0159 vs NT; HIPP: AT8 n=4/5, *p=0.0159 vs NT, unpaired two-tailed t-test). MC1 (CTX: MC1 n=4/5, *p=0.0317 vs NT; HIPP: MC1 n=4/5, *p=0.0159 vs NT, unpaired two-tailed t-test). RZ3 levels are also increased (CTX: RZ3 n=4/5, *p=0.0201 vs NT; HIPP: RZ3 n=4/5, *p=0.0453 vs NT, two-tailed unpaired t-test). h, At variance with HSD (Fig. 1G), hypothermia does not shift tau from soluble to more insoluble fractions. For gel source data see supplementary figure 1. Data are expressed as mean±SEM.

Extended Data Fig. 4:. Effect of L-arginine on p-tau and calpain expression, as well as p-tau in eNOS^−/− mice, calpain and Cdk5 localization, pDARPP-32 with HSD, and IL17 levels

a, Administration of L-arginine (10 g/L in drinking water), starting at week 8 of HSD and continued through week 12, suppresses RZ3 levels in neocortex but not in hippocampus (CTX: RZ3, ND/HSD n=10/10, *p<0.0001 vs ND Veh; HIPP: RZ3, ND/HSD n=10/10, *p=0.0005 vs ND Veh, two-tailed unpaired t-test). b, L-arginine does not affect the increase in serum IL-17 induced by HSD (Veh, ND/HSD n=9/11, *p=0.0002 vs ND Veh; L-Arg, ND/HSD n=9/8, *p<0.0001 vs ND L-Arg, two-tailed unpaired t-test). c, AT8 and RZ3 levels are elevated in the neocortex and the hippocampus of eNOS^−/− mice on ND (AT8: CTX, ND/HSD n=5/4, *p=0.0029 vs WT; HIPP, ND/HSD n=5/4, *p=0.0078 vs WT; RZ3: CTX, ND/HSD n=5/4, *p=0.0003 vs WT; HIPP, ND/HSD n=5/4, *p=0.0128 vs WT, two-tailed unpaired t-test). d, HSD does not increase tau phosphorylation in eNOS^−/− mice (RZ3: HIPP, ND/HSD n=7/8, *p=0.0224 vs ND, two-tailed unpaired t-test). e, Calpain 2 immunoreactivity is present in neuronal cell bodies of the somatosensory and the piriform cortex (size bar=500 μm; 100μm in inset). Representative images from n=3 mice. f, Colocalization of Calpain 2 and Cdk5 in neuronal cell bodies of the piriform cortex (size bar=50 μm; 10 μm in inset). Representative images from n=3 mice. g, HSD has no effect on the phosphorylation of the Cdk5 substrate DARPP-32 in neocortex, ND/HSD n=10/10. h, Administration of the Cdk5 peptide inhibitor TFP5 has no effect on the increase of IL-17 serum levels induced by HSD (Scrambled: ND/HSD n=5/4, *p=0.0002 vs ND Scrambled; TFP5: ND/HSD n=7/8, *p<0.359 vs ND TFP5, two-tailed unpaired t-test). i: L-arginine administration does not alter the levels of calpain 1 and 2 in neocortex and hippocampus, ND/HSD n=3/5. For gel source data see supplementary figure 1. Data are expressed as mean±SEM.

Extended Data Fig. 5:. GSK3β, Pin-1, calpastatin and Cdk5 nitrosylation in HSD, as well as neurovascular coupling, effect of HJ8.8 on p-tau, serum IL-17, and summary diagram.

a, HSD has no effect on the expression and the activity of GSK3β in neocortex, ND/HSD n=10/10. b, HSD did not alter the expression of the prolyl cis/trans isomerase Pin-1, a regulator of tau dephosphorylation, ND/HSD n=5/5. c, The expression of calpastatin, an endogenous inhibitor of calpain activity, is not reduced by HSD, ND/HSD n=10/10. d, Nitrosylation of Cdk5 is reduced in the neocortex of HSD mice (ND/HSD n=9/9, Diet: *p=0.0143; Ascorbate: *p<0.0001, two-way ANOVA and Tukey’s test). e, HJ8.8 reduces AT8 levels in the hippocampus (AT8 – HIPP - IgG: ND/HSD n=13/12; HJ8.8: ND/HSD n=9/13; *p<0.0001, Kruskal-Wallis test and Dunn’s test). RZ3 levels are not altered by HJ8.8. f, Administration of HJ8.8 antibody does alter the increase of IL-17 serum levels induced by HSD (IgG: ND/HSD n=9/9, *p=0.0192 vs ND IgG; HJ8.8: ND/HSD n=7/5, *p=0.0421 vs ND HJ8.8, two-tailed unpaired t-test). g, The CBF increase in somatosensory cortex induced by neural activity evoked by mechanical stimulation of the whiskers is not reduced by HSD in WT, tau^−/− and mice treated with the anti-tau antibody HJ8.8, WT ND/HSD n=5/7, Tau^−/− ND/HSD n=9/8; IgG ND/HSD n=5/5, HJ8.8 ND/HSD n=5/5. h, Western blotting showing enrichment of tau in boiled RIPA neocortical samples (heat stable fraction, HS). Notice that β-actin is lost during the boiling process. Representative images from n=3 experiments. i, Cartoon depicting the mechanisms by which HSD leads to tau phosphorylation and cognitive impairment. HSD elicits a Th17 response in the small intestine, which leads to an increase in circulating IL17. IL17, in turn suppresses endothelial NO production by inducing inhibitory phosphorylation of eNOS at Thr495. The NO deficit results in reduced calpain nitrosylation in neurons, increased calpain activity, p35 to p25 cleavage, activation of Cdk5, and tau phosphorylation, which is ultimately responsible for cognitive dysfunction. In support of this chain of events, rescuing the endothelial NO deficit with L-arginine (L-Arg), lack of tau in tau-null mice, treatment with Cdk5 peptide inhibitor TFP5 or antibodies directed against tau (Tau Ab) prevent the cognitive dysfunction. For gel source data see supplementary figure 1. Data are expressed as mean±SEM.

Fig. 1:. HSD increases tau phosphorylation and insoluble tau.

a, HSD increases AT8 and RZ3 levels. (CTX: AT8, ND/HSD n=8/9, *p<0.0001 vs ND; RZ3, ND/HSD n=12/11, *p<0.0001 vs ND; HIPP: AT8, ND/HSD n=9/9, *p<0.0001 vs ND; RZ3, ND/HSD n=9/9, *p=0.0011 vs ND, two-tailed unpaired t-test). b, HSD increases neuronal AT8 immunoreactivity in the piriform cortex (size bar=500 μm; 100 μm in inset). Representative images from ND and HSD mice (n=5/group). c, Time course of the neocortical increase in AT8 and RZ3 (AT8, 4 weeks: ND/HSD n=4/5, *p=0.0116 vs ND; 8 weeks: ND/HSD n=9/8, *p=0.0066 vs ND; 24 weeks: ND/HSD n=8/9, *p=0.0152 vs ND; 36 weeks: ND/HSD n=4/5, *p=0.0087 vs ND; RZ3, 4 weeks: ND/HSD n=4/5, *p=0.0097 vs ND; RZ3, 8 weeks: ND/HSD n=7/8, *p=0.0084 vs ND; 24 weeks: ND/HSD n=8/9, *p=0.0135 vs ND; 36 weeks: ND/HSD n=4/5, *p=0.0204 vs ND, two-tailed unpaired t-test). d, HSD induces deficits in recognition memory (Diet: *p<0.0001, Time: *p=0.0002; 8 weeks: ND/HSD n=8/11; 12 weeks: ND/HSD n=16/12; 24 weeks: ND/HSD n=14/13 mice/group, two-way ANOVA and Tukey’s test). e, HSD induces deficits in spatial memory (Diet: *p=0.0048, Time: *p<0.0001; ND/HSD n=13/12, two-way RM ANOVA and Bonferroni’s test; Primary Latency Day 5, ND/HSD n=13/12, *p=0.0031 vs ND, two-tailed unpaired t-test). f, Neocortical and hippocampal levels of AT8 correlate with spatial learning impairment (AT8 CTX: BM r=0.4491, *p<0.0001, n=84; NOR r=−0.2621, *p=0.0188, n=80; AT8 HIPP: BM r=0.2073, *p=0.0462, n=93; NOR r=−0.2915, *p=0.0053, n=90, Pearson’s correlation coefficient). g, HSD increases levels of insoluble tau (Western blotting) extracted in RIPA and FA after 12 weeks of treatment (CTX: RIPA, ND/HSD n=7/10, *p=0.0032 vs ND; FA, ND/HSD n=8/7, *p=0.0146 vs ND; HIPP: RIPA, ND/HSD n=7/9, *p=0.0418 vs ND; FA, ND/HSD n=7/9, *p=0.0494 vs ND, two-tailed unpaired t-test). h, HSD increases levels of insoluble tau (electrochemiluminescence method) (CTX: RIPA, ND/HSD n=11, *p=0.0050 vs ND; FA, ND/HSD n=14/11, *p=0.0028 vs ND; HIPP: RIPA, ND/HSD n=6/8, *p=0.0380 vs ND; FA, ND/HSD n=7/8, *p=0.0037 vs ND, two-tailed unpaired t-test). I: HSD shifts tau from the RAB fraction to the less soluble RIPA and FA fractions (CTX: ND/HSD n=9/8, RAB, p=0.4234 vs ND, RIPA, p=0.5414 vs ND, FA, *p=0.0325 vs ND; HIPP: ND/HSD n=5/6, RAB, p=0.2468 vs ND, RIPA, p=0.3290 vs ND, FA, *p=0.0152 vs ND, two-tailed unpaired t-test). For gel source data see supplementary figure 1. Data are expressed as mean±SEM.

Fig. 2:. The NO precursor L-arginine prevents the increase in p-tau induced by HSD.

a–c, Administration of L-arginine (10 g/L in drinking water), starting at week 8 of HSD and continued through week 12, suppresses AT8 accumulation in neocortex and hippocampus (CTX: Veh, ND/HSD n=10/10; L-Arg – ND/HSD n=16/21, *p=0.0045 vs ND; HIPP: Veh, ND/HSD n=10/8; L-Arg – ND/HSD n=7/12, *p=0.0067 vs ND, two-tailed unpaired t-test). d,e, L-arginine treatment improves the cognitive deficits induced by HSD in both the novel object recognition test (Veh – ND/HSD n=12/10, L-Arg – ND/HSD n=6/11; Diet: *p=0.0156, Treatment: *p=0.0406, two-way ANOVA and Tukey’s test) and the Barnes Maze (Primary Latency, Diet: *p=0.0182, Time: *p<0.0001, two-way RM ANOVA plus Tukey’s test; Primary Latency, Day 5, *p=0.0439, Kruskal-Wallis test). For gel source data see supplementary figure 1. Data are expressed as mean±SEM.

Fig. 3:. HSD induces activation of calpain and Cdk5, an effect associated with calpain denitrosylation.

a, HSD did not alter calpain 1 or 2 expression (ND/HSD, n=10), but increased enzyme activity (ND/HSD n=9/9, *p=0.0404 vs ND, two-tailed unpaired t-test). b, HSD increases the cleavage of p35 into p25 (ND/HSD n=8/8, *p=0.0426 vs ND, two-tailed unpaired t-test). c, HSD increases Cdk5 bound to p35p25 (ND/HSD n=10/10, *p=0.0347 vs ND, two-tailed unpaired t-test) and Cdk5 activity (ND/HSD n=10/10, *p=0.0274 vs ND, two-tailed unpaired t-test). d, The Cdk5 peptide inhibitor TFP5 counteracts the HSD-induced increase of AT8 and RZ3 (AT8, CTX: ND/HSD Scrambled n=10/9, ND/HSD TFP5 n=9/9, Diet: *p<0.0001, Treatment: *p=0.0164; AT8, HIPP: ND/HSD Scrambled n=11/10, ND/HSD TFP5 n=10/7, Diet: *p=0.0004, Treatment: *p=0.0360; RZ3, CTX: ND/HSD Scrambled n=10/10, ND/HSD TFP5 n=10/11, Diet: *p<0.0001, Treatment: *p=0.0814; RZ3, HIPP: ND/HSD Scrambled n=12/11, ND/HSD TFP5 n=10/12, Diet: *p<0.0001, Treatment: *p=0.0066, two-way ANOVA plus Tukey’s test). e, TFP5 rescues the spatial memory deficits induced by HSD (Primary Latency, Diet: p=0.6415, Time: *p<0.0001, two-way RM ANOVA plus Tukey’s test; Primary Latency, Day 5, Diet: *p=0.0016, Treatment: p=0.5797, two-way ANOVA plus Tukey’s test). TFP5 also improves cognitive performance of HSD mice at the novel object recognition test (NOR) (Diet: *p=0.0383, Treatment: p=0.1488, two-way ANOVA plus Tukey’s test). f, L-arginine counteracts the increase in calpain (ND/HSD n=8/10, *p=0.0335 vs ND, two-tailed unpaired t-test) and Cdk5 activity induced by HSD and reduces Cdk5 bound to p35p25 (ND/HSD n=7/9, *p=0.0137 vs ND, two-tailed unpaired t-test). g, Calpain 2 nitrosylation is reduced by HSD (ND/HSD n=9/9, Diet: *p=0.0189; Ascorbate: *p<0.0001, two-way ANOVA and Tukey’s test), an effect rescued by L-Arginine (ND/HSD n=6/6, Diet: p=0.9487, Ascorbate: *p<0.0001, two-way ANOVA and Tukey’s test). h, Nitrosylation is suppressed in eNOS^−/− (ND/HSD n=6/6, Genotype: *p=0.0223, Ascorbate: *p=0.0021, two-way ANOVA and Tukey’s test), but not in nNOS^−/− mice (ND/HSD n=4/4, Genotype: p=0.0843, Ascorbate: p<0.0001; two-way ANOVA and Tukey’s test). For gel source data see supplementary figure 1. Data are expressed as mean±SEM.

Fig. 4:. HSD-induced cognitive dysfunction is not observed in tau^−/− mice and is prevented by tau antibodies despite cerebrovascular insufficiency.

a, AT8 and Tau46 are absent in tau^−/− mice in RIPA and heat-stable RIPA fractions. Representative blots from n=5 ND Tau^−/− mice. b, c, HSD does not alter cognition in tau^−/− mice at the NOR (WT: ND/HSD n=9/10; Tau^−/−: ND/HSD n=7/9; Diet: *p=0.0055, Genotype: p=0.1827, two-way ANOVA and Tukey’s test) or the Barnes maze (ND/HSD n=9/10, Diet: p=0.9348, Time: *p<0001, two-way ANOVA and Tukey’s test). d, The CBF increase produced by neocortical application of acetylcholine is reduced in tau^−/− mice (ND/HSD WT, n=6/6, Tau^−/−, n=9/9; Diet: *p<0.0001, Genotype: p=0.7920, two-way ANOVA and Tukey’s test). e, Anti-tau antibodies (HJ8.8, 50mg/Kg/week, i.p.) do not rescue the reduction in resting CBF induced by HSD (IgG: ND/HSD n=11/9; HJ8.8: ND/HSD n=6/5; Diet: *p=0.0061, Treatment: p=0.9367, two-way ANOVA and Tukey’s test). f, HJ8.8 does not rescue the CBF response to acetylcholine (IgG: ND/HSD n=5/5; HJ8.8: ND/HSD n=5/5; Diet: *p=0.0005, Treatment: p=0.8516, two-way ANOVA and Tukey’s test). g, HJ8.8 ameliorates the cognitive dysfunction induced by HSD both at the NOR (IgG: ND/HSD n=15/13; HJ8.8: ND/HSD n=13/15; Diet: *p=0.0001, Treatment: *p=0.0054, two-way ANOVA and Tukey’s test) and the Barnes maze test (Primary Latency - IgG: ND/HSD n=19/15; HJ8.8: ND/HSD n=13/14; Time: *p<0.0001, Diet: *p=0.0358, two-way RM ANOVA and Tukey’s test; Primary Latency - Day 5 - IgG: ND/HSD n=19/16; HJ8.8: ND/HSD n=13/14; *p=0.0202, Kruskal-Wallis and Dunn’s test). For gel source data see supplementary figure 1. Data are expressed as mean±SEM.