A new type of blood-brain barrier aminoacidopathy underlies metabolic microcephaly associated with SLC1A4 mutations

Abstract

Mutations in the SLC1A4 transporter lead to neurodevelopmental impairments, spastic tetraplegia, thin corpus callosum and microcephaly in children. SLC1A4 catalyses obligatory amino acid exchange between neutral amino acids, but the physiopathology of SLC1A4 disease mutations and progressive microcephaly remain unclear. Here, we examined the phenotype and metabolic profile of three Slc1a4 mouse models: a constitutive Slc1a4-knockout mouse; a knock-in mouse with the major human Slc1a4 mutation (Slc1a4-K256E); and a selective knockout of Slc1a4 in brain endothelial cells (Slc1a4tie2-cre). We show that Slc1a4 is a bona fideL-serine transporter at the blood-brain barrier (BBB) and that acute inhibition or deletion of Slc1a4 leads to a decrease in serine influx into the brain. This results in microcephaly associated with decreased L-serine content in the brain, accumulation of atypical and cytotoxic 1-deoxysphingolipids, neurodegeneration, synaptic and mitochondrial abnormalities and behavioural impairments. Prenatal and early postnatal oral administration of L-serine at levels that replenish the serine pool in the brain rescued the observed biochemical and behavioural changes. Administration of L-serine until the second postnatal week also normalized brain weight in Slc1a4-E256K mice. Our observations suggest that the transport of 'non-essential' amino acids from the blood through the BBB is at least as important as that of essential amino acids for brain metabolism and development. We propose that SLC1A4 mutations cause a BBB aminoacidopathy with deficits in serine import across the BBB, required for optimal brain growth, leading to a metabolic microcephaly, which may be amenable to treatment with L-serine.

Keywords: D-serine; mitochondria; mitophagy; serine metabolism; synaptopathy.

Figure 1

Slc1a4-E256K mice exhibit microcephaly and changes in brain amino acids. (A) Representative western blot of Slc1a4 in the brain and other tissues at postnatal Day 11 (P11). β-Actin was used as the loading control. SK = skeletal. (B) Representative western blot of Slc1a4 in the brain of wild-type (WT) and Slc1a4-E256K mice at P11. β-Actin was used as the loading control. The graph shows densitometric quantification normalized by β-actin. Data represent the mean ± standard error of the mean (SEM) of four mice/group. Student’s t-test. (C) In vivo T2-weighted MRI scans of P30 WT and Slc1a4^{tm1e(KOMP)Wtsi} (Slc1a4-KO) mice. (D) Regional brain volumes in 4-week-old WT compared to Slc1a4-E256K mice. Cb = cerebellum; Hp = hippocampus; LV = lateral ventricle; St = striatum. (E) Thickness of corpus callosum (CC) and somatosensory (S1) region of the neocortex (Ctx) in WT compared to Slc1a4-E256K. Data are means ± SEM of six mice/group. Student’s t-test. (F) Brain and body weight of P11 WT and Slc1a4-E256K mice. Data are mean ± SEM of 15 mice/group. Student’s t-test (**P = 0.009; ns = not significant). (G) Amino acid levels in brains of WT (n = 6–9) and Slc1a4-E256K (n = 8–12) P11 mice. ***P < 0.001.

Figure 2

Loss of neuronal and synaptic proteins, degeneration processes and oxidative stress in Slc1a4-E256K mice. (A and B) Western blot of brain tissue from postnatal Day 11 (P11) mice and quantification of proteins involved in neuronal and synaptic function. β-Actin was used as a loading control. Data represent the means ± standard error of the mean (SEM) of 6 mice/group. Student’s t-test. (C) Immunodetection of protein carbonylation levels in wild-type (WT) and Slc1a4-E256K P11 mice. Each lane consists of tissue lysates from a representative mouse. Right: The densitometric quantification of total protein carbonylation normalized to β-actin protein levels. Data represent the means ± SEM of four mice/group. Student’s t-test. (D) Fluoro-jade C (green) and DAPI (blue) staining of WT and Slc1a4-E256K P11 mice, indicating degenerating neurons. Cleaved Parp (cPARP) and phosphorylated Tau (p-Tau) were probed in the same blots. MBP, myelin basic protein. Scale bar = 100 µm ***P < 0.001.

Figure 3

Slc1a4-E256K mice exhibit behavioural impairments and accumulation of toxic 1-deoxysphingolipids. (A) Number ultrasonic vocalizations recorded after pup isolation from the dam and litter from postnatal Day 4 (P4) to P9. Data represent the mean ± standard error of the mean (SEM) of 8–12 mice/group. Student’s t-test followed by Holm-Sidak correction for multiple tests. ***P < 0.001. (B) Classification of ultrasonic call types at P8–P9. Data represent the means ± SEM of 8–12 mice/group. Student’s t-test followed by Holm-Sidak correction for multiple tests. (C) Slc1a4-E256K P11 pups (n = 10) exhibit higher latency in the negative geotaxis than wild-type (WT) (n = 15). Student’s t-test (***P < 0.001). (D) Slc1a4-E256K pups (n = 10) display shorter latency to fall in the front limb suspension test compared with WT (n = 15). Student’s t-test. (E) Slc1a4-E256K pups (n = 10) show unaltered grip strength compared to WT (n = 15). Student’s t-test. (F) Slc1a4-E256K pups (n = 10) and WT (n = 15) display similar latency to fall in the hind limb suspension test. Student’s t-test. (G) Rotor-rod performance of WT and Slc1a4-E256K mice. Student’s t-test. (H) Higher l-alanine to L-serine ratio in the brains of Slc1a4-E256K and Slc1a4^{tm1e(KOMP)Wtsi} (Slc1a4-KO) P11 pups. WT₁ corresponds to littermates of E256K mice, and WT₂ corresponds to littermates of Slc1a4-KO mice. Data are means ± SEM of 9–12 mice/group. Student’s t-test. (I) Sphingolipid synthesis is initiated by the condensation of palmitoyl-CoA with L-serine to form sphinganine (SA) catalysed by serine palmitoyl transferase (SPT). When L-alanine to L-serine ratio increases, SPT uses alanine as substrate, generating atypical sphingolipids derived from 1-deoxysphinganine (1-deoxySA). (J) Increased levels of 1-deoxySA in Slc1a4-E256K and Slc1a4-KO mouse brains. Data are means ± SEM of 15 mice/group at P11. One-way ANOVA revealed a significant effect for group [F(3,55) = 16.42, P < 0.0001], Sidak’s multiple comparisons test (***P < 0.001). (K) Slc1a4-KO mice, but not Slc1a4-E256K mice, show accumulation of 1-deoxysphingosine (1-deoxySO) in the brain. Data are means ± SEM of 15 mice/group at P11. One-way ANOVA revealed a significant effect for the group [F(3,55) = 12.67, P < 0.0001]. Sidak’s multiple comparisons test (***P < 0.001; ns = not significant). (L and M) Levels of the normal SA and SO sphingolipids (C18) were not different in the various groups. One-way ANOVA [F(3,55) = 2.022, P = 0.12; ns = not significant].

Figure 4

Synaptopathy and mitochondrial damage in Slc1a4-E256K mice. (A) Representative electron micrographs of wild-type (WT) and Slc1a4-E256K mice at the molecular layer of the ventral leaflet area of the dentate gyrus demonstrating the presynaptic area (asterisk) in opposition to the postsynaptic area (arrowhead), which contains a distinct postsynaptic density (PSD) and synaptic cleft. Scale bar = 500 nm. (B) Slc1a4-E256K mice display a significant higher number of asymmetric synapses compared to WT mice. Data are from three (WT) and four (Slc1a4-E256K) mice. Student’s t-test. (C) Slc1a4-E256K mice show a decrease in asymmetric presynaptic area and a distribution toward lower values compared to WT mice. Data are from 201 (WT) and 309 (Slc1a4-E256K) synapses. Student’s t-test and Kolmogorov–Smirnov test for cumulative distribution. (D) Slc1a4-E256K mice exhibit a smaller PSD area and a distribution towards lower values. The data are from 176 (WT) and 240 (Slc1a4-E256K) synapses. Student’s t-test and Kolmogorov–Smirnov test for cumulative distribution. (E) The density of symmetric synapses did not differ between the genotypes. ns = not significant. (F) The symmetric synapse bouton area was lower in the Slc1a4-E256K mice, with a distribution skewed to smaller areas. Student’s t-test and Kolmogorov–Smirnov test for cumulative distribution. The data are from 156 (WT) and 230 (Slc1a4-E256K) synapses. (G) Representative electron micrograph of Slc1a4-E256K mice showing mitophagosome-like structures (arrowheads). Scale bar = 500 nm. (H) Quantification of abnormal mitochondria indicates a higher percentage of damaged mitochondria in Slc1a4-E256K mice. Data are from three (WT) and four (Slc1a4-E256K) mice. Student’s t-test. (I) Higher number of mitophagosome-like structures in Slc1a4-E256K mice. (J) Number of mitochondria was unaltered in Slc1a4-E256K mice. (K) Higher levels of autophagy and mitophagy-relevant proteins in Slc1a4-E256K mice. The p62 blot is a re-probe of a beclin1 blot. Student’s t-test (n = 6 mice/group).

Figure 5

Slc1a4 transports L-serine through the blood–brain barrier in vivo and regulates brain L-serine levels. (A–F) Transport through the blood–brain barrier (BBB) in vivo was performed by intracardiac injection of radionuclides. After 10 s, the intravascular radionuclides were flushed by fast perfusion with ice-cold phosphate buffered saline (PBS), and the radioactivity in the brain tissue was monitored to determine their extraction fraction (EF) from the blood. (A) Ratio of EFs of L-[¹⁴C]serine and 3-O-[methyl-³H]D-glucose (3OMG) through the BBB of postnatal Day 11 (P11) mice was lower in Slc1a4-E256K mice brains (n = 7 mice/group). (B) Ratio of L-[¹⁴C]serine and L-[³H]glutamine EFs (n = 6 mice/group) in P11 WT and E256K mice brains. (C) Ratio of t-hydroxy-L-[4-³H]proline (OH-Pro) and L-[¹⁴C]glutamine EFs (WT, n = 9; Slc1a4-E256k mice, n = 8) in P11 WT and E256K mice brains. (D) Ratio of L-[¹⁴C]serine and 3OMG EFs (n = 7 mice/group) in P11 WT and Slc1a4^{tm1e(KOMP)Wtsi} (Slc1a4-KO) mice brain. (E) Ratio of L-[¹⁴C]serine and L-[³H]glutamine EFs in P11 WT and Slc1a4-KO mice brains (WT, n = 10; Slc1a4-KO, n = 8). (F) Ratio of OH-Pro and L-[¹⁴C]glutamine EFs in P11 WT and Slc1a4-KO mice brains (n = 6 mice/group). (G) WT mice (P11) were injected intraperitoneally with saline or L4CIPG (60 mg/kg) and 4 h later, the ratio of OH-Pro and L-[¹⁴C]glutamine EFs were quantified. L4CIPG-treated mice display a decrease in the OH-Pro transport by the BBB in vivo. Brain levels of L-serine (H) and D-serine (I) were quantified by HPLC (n = 5 mice/group) in WT and Slc1a4-KO mice untreated and treated with L4CIPG. Student’s t-test (A–G); one-way ANOVA and Sidak multiple comparisons test (H and I). ***P < 0.001; ns = not significant.

Figure 6

Endothelial cell selective Slc1a4 deletion mice exhibit lower brain weight, motor phenotypes and serine dysregulation. (A) Microvessels (MV) purified from Slc1a4 endothelium-selective deletion (cKO) mice at postnatal Day 11 (P11) exhibited lower Slc1a4 expression when compared to wild-type (WT). By comparison, constitutive Slc1a4^{tm1e(KOMP)Wtsi} (Slc1a4-KO) mice had complete deletion of Slc1a4. (B) cKO mice (P11) exhibited lower brain weight than WT (12 mice/group). The relative weights of the E256K and Slc1a4-KO mice are shown as a percentage of their WT littermates and were calculated from the values in Fig. 1F and Supplementary Fig. 2A. Student’s t-test. (C) cKO mice had normal body weights at P11 (12 mice/group). ns = not significant. (D) cKO had a higher latency to turn in the negative geotaxis (Neg geo) test (12 mice/group). (E) cKO exhibited lower latency to fall in the front limb suspension test (12 mice/group). (F) cKO displayed normal grip strength (12 mice/group). (G) cKO exhibited a lower latency to fall in the hind limb suspension test (12 mice/group). Student’s t-test (D–G). (H) Volcano plot shows changed metabolites in E256K compared with WT. Coloured dots correspond to metabolites with P < 0.05 adjusted by the false discovery rate method of Benjamini, Krieger and Yekutieli (WT, n = 9; E256K, n = 12 mice). γGC = gamma-glutamylcysteine. (I) Significantly changed metabolites in constitutive Slc1a4-KO mice compared to WT (12 mice/group). The metabolites that were changed in both E256K and Slc1a4-KO mice are identified in the graph. (J) Significantly changed metabolites in cKO mice compared to WT (WT, n = 8; cKO, n = 12 mice). The position of serine, the only metabolite changed in all three mouse models, is indicated. (K) Venn diagram indicating the relationship between the changed metabolites in Slc1a4-E256K, Slc1a4-KO and cKO mice brains. (L) Levels of serine by liquid chromatography-mass spectrometry in the three mouse models as a percentage of their WT littermates. Student’s t-test.

Figure 7

Perinatal L-serine administration rescues brain weights, behaviour and neurochemical changes in Slc1a4-E256K. (A and B) The decrease in brain weight observed in both female (A) and male (B) postnatal Day 11 (P11) Slc1a4-E256K mice was prevented by perinatal L-serine supplementation. Data represent the mean ± standard error of the mean (SEM) of 5–12 mice/group. ns = not significant. (C) L-Serine restored the number of ultrasonic vocalizations (USVs) by postnatal Day 10 (P10) Slc1a4-E256K mice to wild-type (WT) levels. Data represent the mean ± SEM of 10–17 mice/group. (D and E) Classification of ultrasonic call types at P10 in naïve mice (D) and mice treated with L-serine (E). Data represent the mean ± SEM of 7–13 mice/group. (F) L-Serine administration rescues negative geotaxis in P11 Slc1a4-E256K mice. Data are mean ± SEM of 10–23 mice/group. (G) Improvement of the front limb suspension test in P11 Slc1a4-E256K mice by L-serine administration. (H) Effect of L-serine administration in the hind limb test. (I) Effect of L-serine administration in the grip strength test. (J) Representative western blots showing the levels of brain NeuN, PSD95, MBP, cPARP, NeuroD6 and pAMPK in L-serine-treated WT and Slc1a4-E256K mice. (K) Densitometric quantification of the proteins in J normalized by β-actin. Data represent the means ± SEM of 5–6 mice/group. (L) Levels of brain L-serine and (M) D-serine in P11 mice with and without perinatal serine administration. Data represent the mean ± SEM of 5–7 mice/group. One-way ANOVA with Bonferroni multi-comparisons test (A–C and F–I) or Sidak’s multiple comparisons test (L and M). Student’s t-test (D and E). NeuN, PSD95, pAMPK and AMPK were sequentially probed in the same blot. cPARP and NeuroD6 were probed sequentially in a different blot. The data for untreated mice in A, B and F–I are the same as in Supplementary Fig. 1 and Fig. 3C–F and were replotted for comparison with L-serine-treated mice.