An intestinal sphingolipid confers intergenerational neuroprotection

Abstract

In animals, maternal diet and environment can influence the health of offspring. Whether and how maternal dietary choice impacts the nervous system across multiple generations is not well understood. Here we show that feeding Caenorhabditis elegans with ursolic acid, a natural plant product, improves axon transport and reduces adult-onset axon fragility intergenerationally. Ursolic acid provides neuroprotection by enhancing maternal provisioning of sphingosine-1-phosphate, a bioactive sphingolipid. Intestine-to-oocyte sphingosine-1-phosphate transfer is required for intergenerational neuroprotection and is dependent on the RME-2 lipoprotein yolk receptor. Sphingosine-1-phosphate acts intergenerationally by upregulating the transcription of the acid ceramidase-1 (asah-1) gene in the intestine. Spatial regulation of sphingolipid metabolism is critical, as inappropriate asah-1 expression in neurons causes developmental axon outgrowth defects. Our results show that sphingolipid homeostasis impacts the development and intergenerational health of the nervous system. The ability of specific lipid metabolites to act as messengers between generations may have broad implications for dietary choice during reproduction.

Fig. 1. UA promotes neuronal health across generations.

a,b, Schematics (top) and fluorescence micrographs (bottom) of posterior lateral mechanosensory left (PLML) anatomy in wild-type (a) and mec-17(ok2109); lon-2(e678) (b) animals expressing the Pmec-4::gfp transgene (zdIs5). Left lateral view, anterior to the left. Scale bars, 25 μm. b, A typical PLM axon break (red line) observed in adult (3-d-old) mec-17(ok2109); lon-2(e678) animals is indicated. c, Chemical structure of UA. d, Adult mec-17(ok2109); lon-2(e678) F₁ animals exposed to UA (50 μM) from P₀ L4 to day 3 of F₁ had reduced PLM axon breaks. e, Timeline of UA exposure. The stages of C. elegans development relevant to UA exposure in the P₀ and F₁ generations are shown (top). Vertical dashed line, demarcation between the P₀ and F₁ generations. f, Continuous UA exposure (P₀ L4 larva to F₁ adult) reduces PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. Specifically, UA exposure of P₀ animals from the L4 larval stage to adult (ii), but not earlier (i) or later ((iii) and (iv)) stages results in a reduction in PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. g, Experimental scheme for the intergenerational inheritance experiment. P₀ animals (L4 larvae) were treated with DMSO (control) or UA for 16 h. Animals of each generation were allowed to lay eggs on untreated plates for 3 h and the adults (3-d-old) were assessed for axon breaks. h, The progeny of mec-17(ok2109); lon-2(e678) P₀ mothers exposed to UA for 16 h (L4 to young adult) have reduced PLM axon breaks for two generations (F₁ and F₂). d,f,h, n = 101 and 103 (d); f, n = 249, 199, 173, 151, 166 and 107 (f); and n = 126, 128, 103, 128, 77 and 97 (h) hermaphrodite animals per condition (left to right). P values were determined using a one-way analysis of variance (ANOVA; f) or unpaired Student’s t-test (d,h). ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Fig. 2. UA induces acid ceramidase (asah-1) expression to promote neuronal health.

a, RNAi-mediated knockdown of rme-2 suppresses UA-induced reduction of PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. b,c, The progeny of P₀ mothers (mec-17(ok2109); lon-2(e678), hrde-1(tm1200); mec-17(ok2109); lon-2(e678) (b) and znfx-1(gg561); mec-17(ok2109); lon-2(e678) (c) animals) exposed to UA for 16 h (L4 to young adult) have reduced PLM axon breaks for two generations (F₁ and F₂). Experimental timing as in Fig. 1g. d, Schematic of transcriptome analysis following UA exposure (top). Total RNA was isolated from L4 larvae exposed to UA for 12 h (compared with the DMSO control). asah-1 was upregulated following UA exposure (Supplementary Table 1). The upregulation of asah-1 messenger RNA in wild-type L4 larvae after 12 h of UA exposure was independently confirmed using qPCR (bottom; n = 3 independent experiments). The housekeeping gene cdc-42 was used as the control. e,f, Loss of asah-1 by RNAi (e) or gene deletion (f) suppresses the ability of UA to reduce PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. g, Representative fluorescence micrographs (top) and calculated levels (bottom) of the Pasah-1::nls::gfp transcriptional reporter. Expression was detected in the intestine from late embryogenesis through to adult (see Extended Data Fig. 4). Pasah-1::nls::gfp expression increased in animals exposed to UA for 12 h compared with the controls (DMSO); a.u. arbitrary units. Lateral views, anterior to the left, of L4 larvae are shown. Pharynx marked by a white asterisk. Scale bar, 25 μm. a–c,e–g, n = 186, 186, 162 and 154 (a); n = 90, 69, 61, 65, 71, 71, 160, 156, 145, 158, 73 and 68 (b); n = 98, 101, 99, 100, 76, 71, 115, 115, 115, 114, 116 and 121 (c); n = 126, 101, 148 and 94 (e); n = 223, 220, 202 and 194 (f); and n = 37 and 43 (g) hermaphrodite animals per condition (left to right). a–g, P values were determined using an ANOVA (a–c,e) or unpaired Student’s t-test (d,f,g). ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Fig. 3. Intestinal and neuronal asah-1 expression has opposing effects on axonal development and health.

a, Expression of asah-1 cDNA driven by the heterologous intestinal promoter (ges-1), but not the hypodermal (dpy-7) or muscle (myo-3) promoters, resulted in reduced PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. b, PLM axon breaks were reduced in mec-17(ok2109); lon-2(e678) animals derived from Pges-1::asah-1 transgenic animals independently of inheritance of the transgene. The mec-17(ok2109); lon-2(e678) animals were either injected with Pges-1::asah-1 (transgenic lines nos. 1 and 2) or uninjected (−). c,d, Overexpression of asah-1 in the nervous system (rab-3 promoter) caused developmental axon outgrowth defects in the mechanosensory neurons of wild-type animals expressing the Pmec-4::gfp transgene (zdIs5). c, Proportion of PLM axon outgrowth defects in Pmec-4::gfp- and Pmec-4::gfp; Prab-3::asah-1-expressing L1 larvae. d, Schematic (top) and fluorescence micrographs (bottom) of ALM (left/right) and PLM (left/right) axons in wild-type (left) and Prab-3::asah-1-expressing animals (right). The typical axon outgrowth defect observed is marked in red. Left lateral view, anterior to the left. Scale bars, 25 μm. a–c, n = 305, 72, 80, 75, 72, 98 and 97 (a); n = 126, 147, 149, 123 and 124 (b); and n = 62 and 95 (c) hermaphrodite animals per condition (left to right). P values were determined using an ANOVA (a) or unpaired Student’s t-test (b,c). Error bars indicate the s.e.m. ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05; and NS, not significant. Source data are provided. Source data

Fig. 4. Intestinal asah-1 neuroprotection requires SphK expression.

a, C. elegans orthologues of de novo- and salvage sphingolipid-pathway components. Sphingolipid metabolic enzymes (nematode and mammalian orthologue) and sphingolipid intermediates are shown. CerS, ceramide synthase; CDase, ceramidase; SPP, S1P phosphatase; and SPL, S1P lyase. b,d,e, The function of SPTL-1, ASAH-1 and SPHK-1 (coloured lettering in a) in PLM axon fragility was examined. Overexpression of sptl-1 (b), asah-1 (d) or sphk-1 (e) cDNA in the intestine (ges-1 promoter) reduces PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. Two independent transgenic lines (nos. 1 and 2, as indicated in the coloured bars) were analysed. c, Exposure of P₀ mothers to ceramide for 16 h (L4 to young adult) results in a reduction of PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. f, RNAi-induced knockdown of sphk-1 suppresses the ability of intestinal asah-1 expression to reduce PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. Transgenic line no. 2 from d was used. b–f, n = 69, 73, 75 and 73 (b); n  = 92, 81 and 88 (c); n  = 76, 72, 79 and 80 (d); n  = 94, 102, 99 and 98 (e); and n  = 99, 97 and 116 (f) hermaphrodite animals per condition (left to right). P values were determined using an ANOVA (c,f) or unpaired Student’s t-test (b,d,e). ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Fig. 5. S1P protects PLM neurons intergenerationally.

a, Chemical structure of S1P. b, Continuous exposure (P₀ L4 larva to F₁ adult) to 20 µM S1P reduces PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. c, PLM axon breaks were reduced in mec-17(ok2109); lon-2(e678) animals when P₀ mothers (L4 to young adult), but not F₁ larvae, were exposed to S1P for 16 h. d, Experimental scheme for the intergenerational inheritance experiment. P₀ animals (L4 larvae) were treated with methanol (control) or S1P for 16 h. Animals of each generation were allowed to lay eggs on untreated plates for 3 h and 3-d-old adults were assessed for axon breaks. e, The progeny of P₀ mothers (mec-17(ok2109); lon-2(e678) animals) exposed to S1P for 16 h (L4 to young adult) had reduced PLM axon breaks for two generations (F₁ and F₂). f, S1P–fluorescein was detected in the oocytes of wild-type adult hermaphrodites after 16 h of feeding. A vehicle-treated control (left) and an animal following S1P–fluorescein treatment (right) are shown. Nomarski micrographs (top) and fluorescence images (bottom) of the same animals are provided. The dashed white lines outline oocytes. Lateral view, anterior to the left. Scale bar, 25 μm. g, RNAi-mediated knockdown of rme-2 suppresses S1P-induced reduction of PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. b,c,e,g, n = 73, 73, 78, 74, 76 and 79 (b); n = 105, 104, 135 and 107 (c); n = 158, 163, 153, 157, 157 and 155 (e); and n = 104, 105, 107 and 104 (g) hermaphrodite animals per condition (left to right). P values were determined using an ANOVA (c,g) or unpaired Student’s t-test (b,e). ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Fig. 6. UA and S1P promote PLM axon transport and microtubule stability.

a,b, Continuous exposure to UA or S1P reduces inappropriate posterior accumulation of UNC-104::GFP (kinesin; a) and mCherry::RAB-3 (pre-synaptic guanosine triphosphatase; b) in the PLM axons of mec-17(ok2109) animals. Scale bars, 25 μm. c,d, Continuous exposure to UA or S1P reduces colchicine-induced PLM axon breaks in mec-17(ok2109); lon-2(e678) (c) and wild-type (d) animals. Wild-type and mec-17(ok2109); lon-2(e678) animals were exposed to 200 µM and 100 µM colchicine, respectively. Two-day-old adult animals were scored. e, Continuous UA or S1P exposure reduces PLM axon breaks in lin-14(n355n679) lon-2(e678) animals. f, Continuous exposure to S1P, but not UA, reduces D-type motor neuron commissure defects in lin-14(n355n679) animals. g, Continuous exposure to UA or S1P reduces PVQ axon defects in ced-10(rp100) animals (g). a–g, Continuous exposure: P₀ to L4 larva to F₁ to adult; n = 245, 247, 103 and 99 (a); n = 107, 119, 96 and 94 (b); n = 112, 152, 146, 99, 101 and 102 (c); n = 76, 74, 96, 76, 75 and 67 (d); n = 149, 175, 138 and 144 (e); n = 109, 106, 97 and 107 (f); and n = 102, 117, 156 and 150 (g) hermaphrodite animals per condition (left to right). P values were determined using an ANOVA (c,d) or unpaired Student’s t-test (a,b,e–g). ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Fig. 7. ASAH-1-mediated neuroprotection requires PBX and MEIS-regulated transcription.

a, PQM-1 and CEH-60 transcription factor ChIP–seq peaks at the asah-1 gene locus (top) aligned with a schematic of the asah-1 endogenous reporter (bottom). The asah-1 upstream region contains binding motifs for CEH-60 and PQM-1 (red line). The location of each binding motif (arrows) is indicated as the distance from the ATG (+1). b, Expression of asah-1::f2a::gfp::h2b in the intestine of an L4 larva (see Extended Data Fig. 7b for other developmental stages). Schematic (top) and overlay of Nomarski and fluorescence image (bottom). Filled circles indicate that expression was detected and open circles that weak/no expression was detected. Lateral view, anterior to the left. Pharynx is marked by a white asterisk; red line indicates bright asah-1::f2a::gfp::h2b expression. Scale bar, 25 μm. c, The expression of asah-1::f2a::gfp::h2b increases following exposure to UA (see Extended Data Fig. 8a for individual intestinal cell measurements). d, Loss of pqm-1 or ceh-60 prevents UA-induced induction of asah-1::f2a::gfp::h2b expression. e–g, Loss of pqm-1 (via RNAi (e) or gene deletion (f)) or ceh-60 (via RNAi (g)) suppresses UA-induced PLM neuroprotection in mec-17(ok2109); lon-2(e678) animals. h, Mutation of the putative PQM-1 binding motif in the asah-1 promoter induces asah-1::f2a::gfp::h2b expression. Int cell pair 1–6, second–seventh pair, respectively, of intestinal cells from the pharynx. i, Mutation of the putative PQM-1 binding motif in the asah-1 promoter reduces PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. j, The progeny of P₀ mothers exposed to S1P for 16 h (L4 to young adult) increases asah-1::f2a::gfp::h2b expression for two generations (F₁ and F₂). c–j, n = 28 and 30 (c); n = 18, 19, 21, 26, 22 and 24 (d); n = 74, 77, 99 and 103 (e); n = 73, 75, 122 and147 (f); n = 65, 71, 150 and 151 (g); n = 32, 30, 32, 30, 32, 30, 32, 30, 32, 30, 32 and 30 (h); n = 100 and 104 (i); and n = 27, 28, 26, 26, 23 and 25 (j) hermaphrodite animals per condition (left to right). P values were determined using an ANOVA (d–g,j) or unpaired Student’s t-test (c,h,i). ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m.; a.u., arbitrary units. Source data are provided. Source data

Extended Data Fig. 1. Natural-product screen for regulators of axon fragility.

a, Natural-product exposure timeline. Upper schematic shows the stages of C. elegans development relevant to continuous natural-product exposure in the P₀ and F₁ generations. b, Quantification of PLM axon breaks in mec-17(ok2109); lon-2(e678) animals following continuous exposure (P₀ to L4 larva to F₁ to adult) of a natural-product library at 50 μM. Orange bars. reduced PLM breaks; red bars, increased PLM breaks; grey bars, not significant. Quantification of PLM axon breaks shown in b, n = 122, 102, 100, 85, 101, 100, 98, 94, 97, 100, 97, 94, 99, 95, 98, 98, 100, 102, 100, 91, 99 and 100 hermaphrodite animals per condition (left to right). P values were determined using an unpaired t-test. ***P ≤ 0.001; *P ≤ 0.05. Error bars indicate the s.e.m. Source data are provided. Source data

Extended Data Fig. 2. Effect of UA on axon fragility and worm growth/behaviour.

a, Continuous exposure (P₀ to L4 larva to F₁ to adult) of UA at 50 or 100 μM, but not 25 μM, rescues PLM axon breaks in mec-17(ok2109); lon-2(e678) animals. b,c, Quantification of locomotion speed of zdIs5 and zdIs5; mec-17(ok2109); lon-2(e678) F₁ animals on day 1 (b) and 3 (c) of adulthood treated with DMSO (control) or UA (P₀ mothers treated for 16 h, L4 to young adult). Data presented as speed in μm s⁻¹ (points). d,e, Quantification of body length of zdIs5 and zdIs5; mec-17(ok2109); lon-2(e678) F₁ animals at day 1 (d) and 3 (e) of adulthood treated with DMSO (control) or UA (P₀ mothers treated for 16 h, L4 to young adult). Data presented as individual lengths in μm (points). f, Exposure of UA (50 μM) from P₀ embryo to F₁ 3-d-old adults (F₁ to A3, continuous) reduces PLM axon breaks in mec-17(ok2109); lon-2(e678) F₁ animals. In contrast, UA exposure (50 μM) from P₀ embryo to P₀ 3-d-old adults (P₀ to A3, continuous), or only during embryogenesis (P₀ to A3 and F₁ to A3, time-specific) does not reduce PLM axon breaks in mec-17(ok2109); lon-2(e678) F₁ animals. Quantification of PLM axon breaks shown in a, n = 101, 102, 101, 103, 101 and 100; f, n = 99, 98, 96, 101, 99, 99, 96 and 97 hermaphrodite animals per condition (left to right). Quantification of locomotion speed shown in b, n = 65 and 40; c, n = 49 and 38 hermaphrodite animals per condition (left to right). Quantification of body length shown in d, n = 66 and 46; e, n = 50 and 50 hermaphrodite animals per condition (left to right). P values were calculated using a one-way analysis of variance (ANOVA; b–e) or unpaired t-test (a,f). ***P ≤ 0.001; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Extended Data Fig. 3. UA RNA-sequencing analysis.

a, Schematic of transcriptome analysis following UA exposure. Total RNA isolated from L4 larvae exposed to DMSO or UA for 12 h; 49 genes were regulated by UA exposure (false discovery rate, 0.02). b, Multidimensional scale plot (MDS) of untreated, DMSO-treated (control) and UA-treated RNA-sequencing samples. Source numerical data are provided in Supplementary Table 1.

Extended Data Fig. 4. asah-1 transcriptional reporter expression analysis.

The Pasah-1::NLS::gfp transcriptional reporter (rpIs165) is expressed in the intestine from the threefold stage of embryogenesis throughout larval development and into adulthood. A schematic of a young adult is shown below the fluorescence micrograph. Note that Pasah-1::NLS::gfp expression is brighter in the anterior intestine (red line). Anterior to the left in larval image. Pharynx marked by a white asterisk. Scale bars, 20 μm.

Extended Data Fig. 5. asah-1 overexpression analysis.

a, Schematic and fluorescence micrographs of ALM and PLM anatomy in L1 larval wild-type (top micrograph) and Pmec-4::asah-1-overexpressing (bottom micrographs) animals expressing the Pmec-4::gfp transgene (zdIs5). Outgrowth defects are marked with red lines and arrows. Scale bar, 20 μm. b,c, Quantification of ALM (b) and PLM (c) outgrowth defects in wild-type, mec-17(ok2109); lon-2(e678); Pmec-4::asah-1, Pmec-4::asah-1 and Prab-3::asah-1 animals. The ALM and PLM outgrowth defects correspond to the micrographs in a. d,e, Schematics and fluorescence micrographs (d) as well as quantification (e) of PVM axon guidance in L4 wild-type and Prab-3::asah-1-overexpressing animals expressing the Pmec-4::gfp transgene (zdIs5). Scale bar, 25 μm. f,g, Schematics (f) and quantification (g) of PVQ axon guidance in L1 wild-type and Prab-3::asah-1-overexpressing animals expressing the Psra-6::gfp transgene (oyIs14). Quantification of ALM outgrowth defects shown in b, n = 60, 42, 39, 46, 48 and 54 hermaphrodite animals per condition (left to right). Quantification of PLM outgrowth defects shown in c, n = 60, 42, 39, 46, 48 and 54 hermaphrodite animals per condition (left to right). Quantification of PVM guidance defects shown in e, n = 45 and 75 hermaphrodite animals per condition (left to right). Quantification of PVQ axon defects shown in g, n = 65 and 73 hermaphrodite animals per condition (left to right). P values were determined using a two-way analysis of variance (ANOVA; b,c) or unpaired t-test (e,g). ****P ≤ 0.0001; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Extended Data Fig. 6. Analysis of sphk-1 function and S1P–fluorescein detection.

a,b, Loss of sphk-1 via RNAi (a) and gene deletion (sphk-1-(ok1097) deletion mutant; b) enhances PLM axon breaks in mec-17(ok2109); lon-2(e678) animals and prevents the protective effect of UA. c, Brightfield and fluorescence image overlay of L4 hermaphrodites exposed to control (methanol) or S1P–fluorescein for 1 h. S1P–fluorescein was detected in the intestinal lumen. Scale bar, 20 μm. Quantification of PLM axon breaks shown in a, n = 99, 97, 128 and 138; b, n = 100, 105, 100 and 100 hermaphrodite animals per condition (left to right). P values were determined using a one-way analysis of variance (ANOVA). **P ≤ 0.01; *P ≤ 0.05; and NS, not significant. Error bars indicate the s.e.m. Source data are provided. Source data

Extended Data Fig. 7. Analysis of the asah-1 endogenous reporter.

a, Nucleotide sequence 200 bp upstream of the asah-1 start codon. Upper sequence shows the wild-type sequence highlighting putative transcription factor-binding motifs for CEH-60 (red) and PQM-1 (blue, motif). The bottom sequence shows the independent mutations introduced to delete each putative transcription factor-binding motif (yellow highlights) that introduced the following restriction enzyme recognition sequences: CEH-60 motif (NotI) and PQM-1 motif (BamHI). b, GFP–H2B expression from the asah-1::f2a::gfp::h2b endogenous reporter (rpIs176) was detected in the intestine from late embryogenesis (threefold stage) throughout larval development and into adulthood. A schematic of each animal is shown next to the fluorescence micrographs. Note that Pasah-1::NLS::gfp expression is consistently brighter in the anterior intestine (red line). Anterior to the left in the worm image. Pharynx marked by a white asterisk. Scale bar, 20 μm.

Extended Data Fig. 8. Regulation of endogenous asah-1 expression.

a, GFP–H2B expression from the asah-1::f2a::gfp::h2b endogenous reporter increases when animals are exposed to UA. Quantification of nuclear fluorescence from each pair of intestinal cells per worm is shown (Int cell pair 1 = second pair of intestinal cells from the pharynx). b, Mutation of putative CEH-60 binding motif in the asah-1 promoter induces GFP–H2B expression from the asah-1 endogenous reporter. Quantification of nuclear fluorescence from each pair of intestinal cells per worm is shown (Int cell pair 1, second pair of intestinal cells from the pharynx). c, Proposed model of UA/S1P intergenerational neuroprotection. UA exposure of P₀ animals (L4 to young adult) elevates intestinal asah-1 expression (probably through inhibition of PQM-1/CEH-60-directed transcriptional repression) and increases S1P levels. S1P is transferred within intestinally derived yolk to oocytes via the RME-2 receptor and promotes axon health in these animals (top). In F₁ progeny, elevated S1P is detected in the intestine, which causes asah-1 expression to increase and generate additional S1P that is transmitted, via the yolk, to F₂ oocytes and promotes axon health in these animals (bottom). UA/S1P neuroprotection is not observed in F₃ progeny, probably due to reduced S1P levels that are below the threshold required for neuroprotection. Quantification of asah-1::f2a::gfp::h2b fluorescence shown in a, n = 28, 29, 28, 29, 28, 29, 28, 29, 28, 29, 28 and 29; b, n = 29 animals per condition (left to right). P values were determined using and unpaired t-test. ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; *P ≤ 0.05. Error bars indicate the s.e.m. Source data are provided. Source data