Blood factors transfer beneficial effects of exercise on neurogenesis and cognition to the aged brain

Abstract

Reversing brain aging may be possible through systemic interventions such as exercise. We found that administration of circulating blood factors in plasma from exercised aged mice transferred the effects of exercise on adult neurogenesis and cognition to sedentary aged mice. Plasma concentrations of glycosylphosphatidylinositol (GPI)-specific phospholipase D1 (Gpld1), a GPI-degrading enzyme derived from liver, were found to increase after exercise and to correlate with improved cognitive function in aged mice, and concentrations of Gpld1 in blood were increased in active, healthy elderly humans. Increasing systemic concentrations of Gpld1 in aged mice ameliorated age-related regenerative and cognitive impairments by altering signaling cascades downstream of GPI-anchored substrate cleavage. We thus identify a liver-to-brain axis by which blood factors can transfer the benefits of exercise in old age.

Conflict of interest statement

Competing interests: The authors declare no conflict of interest. A.M.H., X.F., and S.A.V. are named as inventors on a patent application arising from this work.

Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.

Figures

Fig. 1.. Systemic administration of exercise-induced circulatory blood factors ameliorates impaired neurogenesis and cognition in the aged hippocampus.

(A) Plasma was collected from exercised or sedentary aged (18 months) mice and administered to sedentary aged mice 8 times over 24 days (100 μl per intravenous injection). Schematic illustrates chronological order of plasma administration from exercised aged mice and cognitive testing. (B) Representative microscopic fields and quantification of GFAP/Sox2 double-positive, Dcx-positive, and NeuN/BrdU double-positive cells in the dentate gyrus (DG) of the hippocampus of naïve aged mice administered plasma from sedentary (Sed) or exercised (Run) aged mice (n = 10 or 11 per group; arrowheads point to individual cells; scale bar, 100 μm). Dapi, 4′,6-diamidino-2-phenylindole. (C) Western blot and quantification of BDNF in the hippocampus of naïve aged mice administered plasma from sedentary or exercised aged mice (n = 6 to 10 per group). Quantification is normalized to β-tubulin. (D and E) Spatial learning and memory were assessed by RAWM as the number of entry errors committed during the training and testing phases. Overall learning and memory were analyzed between block 1 and block 10 (1 block = 3 trials; n = 12 to 15 per group). (F to H) Associative fear memory was assessed using contextual (G) and cued (H) fear conditioning as percent time spent freezing 24 hours after training. Baseline freezing (F) was assessed as the percentage of time spent freezing prior to fear conditioning (n = 12 to 19 per group). Data are means ± SEM; *P < 0.05, **P < 0.01, ****P < 0.0001 [t test in (B), (C), (F), (G), and (H); repeated-measures analysis of variance (ANOVA) with Bonferroni post hoc test in (D); ANOVA with Tukey post hoc test in (E)].

Fig. 2.. Exercise increases systemic levels of Gpld1 in mature and aged mice and healthy elderly humans.

(A) Venn diagram of results from proteomic screens of aged (18 months) and mature (7 months) exercised mice. Numbers of proteins whose concentrations increase with exercise in aged and mature mice are shown in the blue and teal regions, respectively. Proteins common to both groups are listed at the right. (B) Enrichment analysis of the 12 proteins up-regulated by exercise in mature and aged mice. Gpld1 and Pon1 are listed next to the processes in which they are implicated. Numerals at far right are numbers of proteins represented in each process. (C and D) Western blots with corresponding Ponceau S stains and quantification of Gpld1 in equal volumes of blood plasma from individual aged (C) and mature (D) sedentary and exercised mice (n = 4 or 5 per group). (E and F) Correlation between plasma Gpld1 levels in exercised and sedentary aged mice and number of errors committed during the final block of RAWM (E) or time spent freezing in contextual fear conditioning (F). (G) Western blot and quantification of Gpld1 in equal volumes of blood plasma from individual sedentary (<7100 steps per day) and active (>7100 steps per day) healthy elderly humans (aged 66 to 78 years; n = 8 to 12 per group). Data are means ± SEM; *P < 0.05 [t test in (C), (D), and (G); linear regression in (E) and (F)].

Fig. 3.. Increased systemic GPLD1 ameliorates impaired neurogenesis and cognition in the aged hippocampus.

(A and B) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) of Gpld1 across tissues in sedentary aged mice (A) and in liver of exercised and sedentary aged mice (B). Gene expression is measured relative to Gapdh (n = 5 or 6 per group). Abbreviations: Cbm, cerebellum; Ctx, cortex; Hipp, hippocampus; TA, tibialis anterior muscle. (C) Aged (18 months) mice were given HDTVI of expression constructs encoding either Gpld1 or GFP control. Schematic illustrates chronological order of HDTVI, cognitive testing, and cellular and molecular analysis. (D) qRT-PCR of Gpld1 in liver of aged mice expressing Gpld1 or GFP control. Gene expression is measured relative to Gapdh (n = 5 per group). (E) Western blot with corresponding Ponceau S stain and quantification of Gpld1 in equal volumes of blood plasma from individual aged mice expressing Gpld1 or GFP control (n = 4 per group). (F) Representative microscopic fields and quantification of GFAP/Sox2 double-positive, Dcx-positive, and NeuN/BrdU double-positive cells in the DG of the hippocampus of aged mice expressing Gpld1 or GFP control (n = 6 per group; arrowheads point to individual cells; scale bar, 100 μm). (G) Western blot and quantification of BDNF in the hippocampus of aged mice expressing Gpld1 or GFP control (n = 6 per group). Quantification is normalized to β-tubulin. (H and I) Spatial learning and memory were assessed by RAWM as number of entry errors committed during the training and testing phases. Overall learning and memory was analyzed between block 1 and block 10 (1 block = 3 trials; n = 26 per group). (J) Spatial working memory was assessed by YMaze as time spent in the start, trained, and novel arms during the testing phase (n = 23 to 25 per group). (K) Object recognition memory was assessed by NOR as time spent exploring a novel object 24 hours after training (n = 8 to 12 per group). Data are means ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 [t test in (B), (D), (E), (F), (G), and (K); repeated-measures ANOVA with Bonferroni post hoc test in (H); ANOVA with Tukey post hoc test in (I) and (J); one-sample t test versus 50% in (K)].

Fig. 4.. Increased systemic Gpld1 alters signaling cascades downstream of GPI-anchored substrate cleavage in the aging systemic milieu.

(A) Heat maps of top 20 proteins up- and down-regulated in blood plasma of aged mice after Gpld1 HDTVI relative to GFP HDTVI control, identified by mass spectrometry. (B and C) Western blot with corresponding Ponceau S stain and quantification of plasminogen [Plg; (B)] and vitronectin [Vtn; (C)] in equal volumes of blood plasma from individual aged mice 24 hours after HDTVI of Gpld1 or GFP control (n = 4 per group). (D) List of 41 proteins down-regulated in blood plasma from aged mice after exercise. (E and F) Enrichment analysis of plasma proteins down-regulated with Gpld1 HDTVI (E) or exercise (F) in aged mice, as identified by mass spectrometry. The number of proteins represented in each process is listed to the right of each bar. Data are means ± SEM; **P < 0.01, ***P < 0.001 [t test in (B) and (C)].

Fig. 5.. GPI-anchored substrate cleavage is associated with restorative effects of Gpld1 on the aged hippocampus.

(A) Luminescence-based quantification of alkaline phosphatase activity in cell culture supernatant 48 hours after transfection with ubiquitin-lox-stop-lox-PLAP (GPI-anchored alkaline phosphatase) and EF1a-Cre, in combination with GFP, Gpld1, H133N-Gpld1, or H158N-Gpld1 (n = 3 samples per group). (B) Aged (18 months) mice were given HDTVI of expression constructs encoding Gpld1, catalytically inactive H133N-Gpld1, or GFP control. Schematic illustrates chronological order of HDTVI, cognitive testing, and cellular and molecular analysis. (C) Western blot with corresponding Ponceau S stain and quantification of Gpld1 in equal volumes of blood plasma from individual aged mice expressing Gpld1, Gpld1-H133N, or GFP control (n = 6 per group). (D) Representative microscopic fields and quantification of GFAP/Sox2 double-positive, Dcx-positive, and NeuN/BrdU double-positive cells in the DG of the hippocampus of aged mice expressing Gpld1, Gpld1-H133N, or GFP control (n = 7 per group; arrowheads point to individual cells; scale bar, 100 μm). (E) Western blot and quantification of BDNF in the hippocampus of aged mice expressing Gpld1, Gpld1-H133N, or GFP control (n = 8 per group). Quantification is normalized to β-tubulin. (F) Object recognition memory was assessed by NOR as time spent exploring a novel object 24 hours after training (n = 9 to 11 per group). (G and H) Spatial learning and memory were assessed by RAWM as number of entry errors committed during the training and testing phases. Overall learning and memory were analyzed between block 1 and block 10 (1 block = 3 trials; n = 12 to 14 per group). Data are means ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 [repeated-measures ANOVA with Bonferroni post hoc test in (G); ANOVA with Tukey post hoc test in (C), (F), (G), and (H); one-sample t test versus 50% in (F)].