Abstract

   Transmission and secretion of signals via the choroid plexus (ChP)
   brain barrier can modulate brain states via regulation of cerebrospinal
   fluid (CSF) composition. Here, we developed a platform to analyze
   diurnal variations in male mouse ChP and CSF. Ribosome profiling of ChP
   epithelial cells revealed diurnal translatome differences in metabolic
   machinery, secreted proteins, and barrier components. Using ChP and CSF
   metabolomics and blood-CSF barrier analyses, we observed diurnal
   changes in metabolites and cellular junctions. We then focused on
   transthyretin (TTR), a diurnally regulated thyroid hormone chaperone
   secreted by the ChP. Diurnal variation in ChP TTR depended on Bmal1
   clock gene expression. We achieved real-time tracking of CSF-TTR in
   awake Ttr^mNeonGreen mice via multi-day intracerebroventricular fiber
   photometry. Diurnal changes in ChP and CSF TTR levels correlated with
   CSF thyroid hormone levels. These datasets highlight an integrated
   platform for investigating diurnal control of brain states by the ChP
   and CSF.

   Subject terms: Circadian regulation, Blood-brain barrier, Molecular
   neuroscience
     __________________________________________________________________

   The choroid plexus (ChP) modulates cerebrospinal fluid (CSF)
   composition and the blood-CSF barrier. Here the authors show that the
   ChP is a critical circadian component with time-of-day variations in
   translation, barrier, and metabolism to alter CSF composition.

Introduction

   Cerebrospinal fluid (CSF) plays critical roles in regulating the
   central nervous system (CNS) throughout life, including the
   distribution of essential health and growth-promoting factors and
   clearing of waste from the brain^[68]1. CSF production robustly varies
   across the 24-h light–dark cycle^[69]2. CSF distribution between brain
   parenchyma interstitial fluid, ventricles, and cervical lymph nodes
   also varies across the 24-h day^[70]3–[71]5. Over a century of studies
   suggest that the CSF contains biomarkers for circadian rhythmicity and
   plays key roles in relaying the output of diurnal clocks to target
   brain tissues. Early studies showed that CSF can carry circadian cues
   for drowsiness^[72]6,[73]7 and satiety^[74]8, and experimental models
   indicate that diffusible factors released from the suprachiasmatic
   nucleus (SCN) of the hypothalamus into the CSF can mediate the
   circadian rhythmicity of locomotion^[75]9–[76]13. The brain’s SCN
   “master clock” is entrained by daily light–dark cycles and in a day of
   12 h of light and 12 h of dark, C57BL/6 and CD1 laboratory mice are
   primarily active in the first half of the dark phase. However, progress
   toward understanding diurnal CSF regulation has been hampered by a
   limited understanding of how tissues that contribute to CSF govern its
   composition and a lack of specialized tools for tracking these changes
   in vivo or in real time.

   The choroid plexus (ChP) is a key source of CSF. The specialized ChP
   epithelial cells that directly modulate CSF contents also display
   cell-autonomous circadian rhythmicity in gene expression of core
   components of the molecular clock^[77]14–[78]17. These molecular
   circadian clocks depend on transcriptional-translational feedback loops
   as well as epigenetic, translational, and post-translational mechanisms
   that interact to enable both a robust and responsive
   clock^[79]18–[80]21. In mammals, the molecular clock is a negative
   feedback loop in which a heterodimer of the master circadian regulator
   BMAL1 and its partners CLOCK (or NPAS2) activate transcription of
   clock-controlled genes including Period (Per) and Cryptochrome (Cry)
   that code for repressors of BMAL1 heterodimer activity, thus closing
   the loop that generates rhythms of approximately 24 h^[81]22,[82]23. In
   addition to transcription, ribosome biogenesis and translation have
   emerged as direct outputs of the circadian molecular clock at the
   cellular level in synchronized cell lines and in the
   liver^[83]24–[84]26. In fact, BMAL1 directly associates with
   translational machinery to promote protein synthesis^[85]25. While the
   circadian adaptations of each tissue involve both transcriptional and
   translational regulation, essentially nothing is known about the
   regulation of translation at the blood-CSF barrier.

   Here, we developed an experimental platform to analyze the rhythmicity
   of ChP and CSF with respect to circadian clock gene expression and its
   dependence on external cues. We adapted tools to evaluate broad diurnal
   changes in transcript abundance, transcript ribosomal association, and
   protein levels. Using these tools, we pinpointed key pathways that
   change throughout the day and generated baseline datasets from unique,
   limited tissue to describe diurnal variations in the ChP translatome,
   secretome, and metabolome. These observations demonstrated widespread
   diurnal regulation of ChP secretion and barrier function across hours
   during the circadian day and in response to feeding cues, resulting in
   changes in CSF composition. We then established an intravital CSF fiber
   photometry system to track ChP output. Together, these multi-modal
   results illustrate the utility of our integrated platform for tracking
   diurnal dynamics and functions of the ChP-CSF system and raise testable
   hypotheses about how the ChP-CSF axis forms an important bridge between
   body and brain rhythmicity.

Results

Choroid plexus translation is diurnally regulated

   To investigate whether translation and protein synthesis differ in ChP
   between light and dark phases, we monitored the mTOR-effector kinase
   ribosomal protein S6 kinase 1 (S6K1), which drives translation when
   phosphorylated. Using phosphorylation of the S6K target ribosomal
   protein S6 as a measure of mTOR pathway activity and general
   translation capacity, we observed increased phosphorylation of
   ribosomal protein S6 on Ser 240/244 (pS6) at 9 p.m. during the dark
   phase compared to 9 a.m. during the light phase (Fig. [86]1a–c). Higher
   pS6 at 9 p.m. suggests that increased levels of active translation
   machinery, and therefore higher protein synthesis capacity, occur
   during the dark phase relative to the light phase in mouse ChP. The
   circadian rhythmicity of ChP core molecular clock component transcripts
   Bmal1 and Per2 was in phase with that of the liver (Fig. [87]1d,
   Supplementary Fig. [88]1a–e). The liver served as an internal control
   since circadian changes in translation and metabolism have been
   robustly defined in the liver. The times of day with differential ChP
   pS6 expression (low at 9 a.m. and high at 9 p.m.) were inversely
   correlated with phasic Bmal1 mRNA expression. The relative increase in
   dark phase pS6 depended on Bmal1, as differential S6 phosphorylation
   was abrogated in ChP from Bmal1-knockout mice (Fig. [89]1e, f). Using
   ribosome biogenesis as another readout of increased ChP translation
   capacity, we found that nucleolar volume as detected by the nucleolar
   protein Fibrillarin was higher at 9 p.m. than at 9 a.m. (Fig. [90]1g,
   h), consistent with reports in liver that the circadian clock
   coordinates ribosome biogenesis^[91]24. We next used
   O-propargyl-puromycin (OPP)^[92]27 delivered by intraperitoneal
   injection to visualize actively elongating nascent polypeptides in vivo
   at the cellular level. OPP incorporation was higher at 9 p.m. compared
   to 9 a.m. in ChP epithelial cells (Fig. [93]1i), consistent with their
   larger nucleolar volumes and higher pS6 levels at night. Overall,
   translation was increased during the dark phase and diurnal variation
   in translation was dependent on systemic Bmal1 expression.

Fig. 1. Choroid plexus translation is diurnally regulated to be higher during
the dark phase.

   [94]Fig. 1
   [95]Open in a new tab

   a Immunoblotting of ChP protein extracts showed increased S6
   phosphorylation (pS6) relative to total S6 at 9 p.m. (blue) compared to
   9 a.m (orange). All vignettes were cropped from the same membrane. b
   Ratio of pS6 to total S6 immunoblotting at 9 a.m. (orange) and 9 p.m.
   (blue). *p < 0.05 (p = 0.038) Student’s two-tailed unpaired t test,
   N = 9 biologically independent animals at each time over 3 independent
   experiments. Data are presented as mean values ± standard deviation
   (SD). c Immunostaining for pS6 in lateral ventricle (LV) ChP also shows
   increased pS6 at 9 p.m.; scale bar = 50 μm. d RT-qPCR analysis of Bmal1
   expression in LV ChP (blue) and liver (red) showed cycling of the
   molecular clock and revealed similar phase between the two tissues.
   Data are presented as mean values ± standard deviation (SD). e
   Immunoblotting of ChP protein extracts for pS6 showed that increased S6
   phosphorylation at 9 p.m. is dependent on Bmal1. f Ratio of pS6 to
   total S6 from immunoblots of ChP from Bmal1-null vs. WT mice.
   **p < 0.01 (p = 0.0017) Student’s two-tailed unpaired t test, N = 8
   biologically independent animals at each time over 4 independent
   experiments. Data are presented as mean values ± standard deviation
   (SD). g Immunohistochemistry of the nucleolar protein Fibrillarin (red)
   in ChP epithelial cells showed larger nucleoli at 9 p.m., during the
   dark phase; scale bar = 5 μm. N = 4 at each time. h Quantification of
   median nucleolar volume. *p < 0.05 Student’s two-tailed unpaired t
   test, N = 5 biologically independent animals at each time over 2
   independent experiments. Data are presented as mean values of the
   median of 50 nucleoli per animal ± standard error of the mean (SEM). i
   OPP incorporation assay in adult ChP epithelial cells showed an
   increased rate of protein synthesis at 9 p.m. than at 9 a.m.; scale
   bar = 100 μm. j RPL10A-conjugated EGFP expression in ChP epithelial
   cells after Foxj1-Cre recombination in TRAP-BAC mice; scale
   bar = 100 μm. k Heatmaps and hierarchical clustering of transcripts
   associated with RPL10A vs. those in the supernatant at either 9 a.m. or
   9 p.m. (adjusted p < 0.05). l Number of distinct, unique protein-coding
   genes on (solid) or off (transparent) the RPL10A ribosomal subunit at
   9 a.m. (orange) and 9 p.m. (blue) (adjusted p < 0.05). m Average FPKM
   of protein-coding genes on (solid) or off (transparent) the RPL10A
   ribosomal subunit at 9 a.m. (orange) and 9 p.m. (blue) (adjusted
   p < 0.05). N = 3 biologically independent samples (LV ChP pooled from
   three animals per sample) at each time in 1 experiment. Data are
   presented as mean values ± standard deviation (SD). Male mice were
   analyzed. Source data are provided as a Source Data file (Source Data).

   We next used ribosome profiling to determine whether overall levels of
   translation in ChP and the number of unique transcripts associated with
   ribosomes were increased during the dark phase (9 p.m.). Using
   Translating Ribosomal Affinity Purification (TRAP)^[96]28,[97]29 from
   ChP at 9 a.m. and 9 p.m., we identified transcripts that were
   prioritized for translation by the subset of polysome-enriched
   ribosomes containing large ribosomal subunit protein RPL10A^[98]30. ChP
   epithelial cells were targeted by crossing FoxJ1:cre mice^[99]31 with
   floxed TRAP (EGFP:L10a) mice^[100]29,[101]32 (Fig. [102]1j), and mRNA
   associated with the large ribosomal subunit RPL10A was purified for
   sequencing from dissected lateral ventricle (LV) ChP (cre-negative
   animals were used as controls). TRAP analyses revealed unique
   transcripts differentially associated with RPL10A at 9 a.m. vs. 9 p.m.
   and those transcripts preferentially in the unbound supernatant
   (Fig. [103]1k). While both times had more unique transcripts in the
   supernatant than associated with RPL10A (Fig. [104]1l), both the total
   number of unique transcripts and the normalized number of reads (FPKM;
   Fragments Per Kilobase of transcript per Million mapped reads) of those
   transcripts associated with the ribosomes were higher during the dark
   phase (9 p.m.), consistent with a more diverse translatome and larger
   overall translation output at 9 p.m. during the dark phase in LV ChP
   than during the light phase at 9 a.m. (Fig. [105]1l, m). Together, this
   analysis of ChP translation and protein synthesis revealed meaningful
   differences between the dark and light phases including higher levels
   of pS6 and protein synthesis during the dark phase. Additionally, we
   demonstrated that Bmal1 was required for diurnal differences in
   phosphorylation of the ribosomal protein S6 in ChP (Fig. [106]1).

   Next, we tested the differential association of specific transcripts
   with the ribosome between light and dark phases in ChP epithelial
   cells. Analyses of all ChP epithelial cell transcripts associated with
   ribosomes revealed 779 differentially translated transcripts at 9 a.m.
   vs. 9 p.m.: 431 enriched at 9 a.m., and 348 enriched at 9 p.m.
   (Fig. [107]2a). Gene sets identified from differentially enriched genes
   were substantially different between the two times (Fig. [108]2b–d;
   Supplementary Fig. [109]2a). Relevant to the role of ChP as a secretory
   epithelium, secreted proteins (Fig. [110]2e) and signal
   peptide-containing transcripts without a transmembrane domain, were
   differentially diurnally regulated in ChP (Supplementary
   Fig. [111]2b–g). Gene sets associated with translation and active
   mitochondria were enriched at 9 p.m. (Fig. [112]2f, g), consistent with
   the abovementioned translation changes and crosstalk between metabolism
   and circadian clocks^[113]33,[114]34. These data raise the hypothesis
   that ChP metabolism responds to environmental cues and correlates with
   increased translation activity. We also identified diurnal changes in
   ChP gene sets associated with barrier adhesion and permeability. These
   transcripts were largely upregulated at 9 a.m. (Fig. [115]2h),
   suggesting that the ChP epithelial barrier changes throughout the day,
   consistent with diurnal variations in ChP function. Taken together,
   these results show that the ChP translatome and secretome are regulated
   diurnally, with differential effects on translational machinery,
   secreted proteins, metabolism, and barrier components.

Fig. 2. Association of ChP cytoplasmic, membrane bound, and secreted protein
mRNAs with the ribosome is diurnally regulated.

   [116]Fig. 2
   [117]Open in a new tab

   a Heatmaps and hierarchical clustering of z-scores for transcripts
   associated with RPL10A in LV ChP at 9 a.m. vs. 9 p.m. (adjusted
   p < 0.05). CuffDiff; adjusted p < 0.05; |log2 FC|>0.4. b All
   significantly regulated pathways from pathway enrichment and
   overrepresentation analysis (corrected for false discovery rate (FDR).
   c, d Top 7 enriched functional annotation clusters by DAVID in LV ChP
   at 9 a.m. (orange) and 9 p.m. (blue). e Volcano plot of significantly
   enriched RPL10A-associated transcripts with a product predicted to be
   secreted from LV ChP at 9 a.m. (orange) and 9 p.m. (blue). *p ≤ 0.01;
   **p ≤ 0.001; ***p ≤ 0.0001; ****p ≤ 0.00001; adjusted p values. f TRAP
   data for the individual genes associated with ribosomes and
   mitoribosomes. *p ≤ 0.01; **p ≤ 0.001; adjusted p values. g TRAP data
   for the individual genes associated with oxidative phosphorylation.
   *p ≤ 0.01; **p ≤ 0.001; ***p ≤ 0.0001; ****p ≤ 0.00001; adjusted p
   values. h TRAP data for the enriched pathways associated with barrier
   permeability. The median log[2] fold change value is indicated by the
   solid vertical bar. *p ≤ 0.01; **p ≤ 0.001; ***p ≤ 0.0001;
   ****p ≤ 0.00001; adjusted p values. Male mice were analyzed. Source
   data are provided as a Source Data file (Source Data).

TTR accumulates in the ChP during the dark phase and depends on Bmal1 and
feeding cues

   Of the top 5 dark phase-secreted proteins identified to be diurnally
   regulated at the level of translation, the mostly highly expressed gene
   was Transthyretin (Ttr) (Fig. [118]3a). Ttr is a signature gene of the
   ChP that encodes a carrier protein for thyroid hormone, amyloid-β, and
   retinoic acid^[119]35–[120]37. We validated the TTR expression pattern
   by immunoblotting at 9 p.m. vs. 9 a.m. (Fig. [121]3b, Supplementary
   Fig. [122]3c). A recent study in rat ChP identified about 1.5× rhythmic
   variation of Ttr transcript^[123]38, where the nadir and zenith were
   equivalent to noon and midnight, respectively. However, in our
   experiments in mouse, the modulation of Ttr expression was not
   reflected at the level of transcript abundance (Supplementary
   Fig. [124]3a, b, i). Rather our data indicate that ribosomal
   association alone is likely to account for diurnal Ttr expression
   modulation (Fig. [125]3a).

Fig. 3. Ttr is preferentially translated by the ChP during the dark phase and
is dependent on feeding.

   [126]Fig. 3
   [127]Open in a new tab

   a The top 5 TRAP candidates enriched at 9 p.m. from ChP TRAP pulldown.
   FPKM for each individual mouse is shown at 9 a.m. (orange circles) and
   9 p.m. (blue squares) for those transcripts associated with RPL10A
   (solid circles) and those in the supernatant (empty circles). These are
   candidates for preferential regulation at the level of translation.
   N = 3 biologically independent samples (LV ChP pooled from 3 animals
   per sample). Data are presented as mean values ± SEM. b Immunoblotting
   analysis and quantification of LV ChP expression of TTR protein.
   *p < 0.05 (p = 0.0475); Student’s two-tailed unpaired t test, N = LV
   ChP from 10 biologically independent animals at each time over 3
   independent experiments. Data are presented as mean values ± standard
   deviation (SD). c Immunoblotting of ChP protein extracts for TTR showed
   that increased protein levels at 9 p.m. is dependent on Bmal1.
   Quantification of the TTR intensity ratio between 9 a.m. and 9 p.m. in
   WT and Bmal1^−/− animals. *p < 0.05 (p = 0.0118); Student’s two-tailed
   unpaired t test, N = 8 ratios from biologically independent pairs of
   animals from each genotype over 3 independent experiments. Data are
   presented as mean values ± standard deviation (SD). d Schematics of the
   restricted feeding regimes used as related to the ad libitum paradigm
   that was used for the previous studies up until this point.
   Immunoblotting of ChP protein extracts for TTR showed that increased
   protein levels at 9 p.m. is dependent on feeding behavior. e
   Experimental setup for explant studies including PER2:LUC luminometry
   and TTR:mNeonGreen microscopy. f Representative PER2:LUC oscillations
   in isolated culture before and after addition of dexamethasone. g
   Average period length for PER2:LUC ChP oscillations in isolated culture
   before and after dexamethasone are close to 24 h. N = 3 ChP over 2
   independent experiments. Data are presented as mean values ± SEM. h
   Genetic targeting used to generate Ttr^mNeonGreen mouse showing
   appropriate distribution of mNeonGreen to ChP epitheial cells. Scale
   bar = 100 μm; inset scale bar = 50 μm. i Explant preparation from
   Ttr^mNeonGreen ChP (scale bar = 100 μm; inset scale bar = 100 μm) shows
   j rhythmic oscillations of mNeonGreen across the day ex vivo. N = 2 LV
   ChP over 2 independent experiments. Data are presented as mean values
   normalized to the final value and shaded area represents range. k
   Immunoblotting of TTR in LV ChP across a whole day at 3-h intervals
   shows sharp upregulation of TTR in ChP during the dark phase.
   *p = 0.034 corrected for 8 comparisons with Šídák’s multiple
   comparisons test. Solid line represents average (normalized to vinculin
   and TTR average value) and shaded area represents standard deviation
   (SD). N = 3 biologically independent animals at each time across 3
   independent experiments. l Immunoblotting of constant volumes of CSF
   shows sharp upregulation of TTR in CSF at 10 p.m. and 11 p.m. relative
   to transferrin (Tf). *p < 0.05 (p = 0.0437), *p < 0.01 (p = 0.00315);
   Student’s two-tailed unpaired t test. N = 3 biologically independent
   animals at each time across 2 independent experiments. Data are
   presented as mean values ± standard deviation (SD). m CSF concentration
   of active thyroid hormone T3 (triiodothyronine) and the circulating
   form T4 (thyroxine) at 5 p.m. (orange) and 11 p.m. (blue). *p < 0.05
   (p = 0.0369), Student’s two-tailed unpaired t test. N = 7 biologically
   independent animals at each time in one experiment. Data are presented
   as mean values ± standard deviation (SD). Male mice were analyzed.
   Source data are provided as a Source Data file (Source Data).

   To uncover how ChP-TTR is regulated throughout the day, we first used
   Bmal1-null mice with a disrupted molecular clock. Diurnal ChP-TTR
   protein expression depended on the molecular clock and was altered in
   ChP from Bmal1-null mice (Fig. [128]3c). Since the intrinsic
   cell-autonomous clock is absent in Bmal1-null mice, all circadian
   behaviors, including feeding, are disrupted^[129]39,[130]40. Therefore,
   this model could not discern whether the observed ChP molecular clock
   was entrained by pacemaker-driven intrinsic cues alone or secondarily
   to feeding or other nutrition-dependent behaviors. To distinguish
   between these variables, we took advantage of the fact that feeding
   cues have the ability to entrain many peripheral clocks^[131]34 and
   tested the in vivo effects of switching from an ad libitum feeding
   paradigm to time-restricted feeding (Fig. [132]3d; Supplementary
   Fig. [133]3e–h). This feeding restriction paradigm allowed us to
   decouple feeding-driven metabolic and behavioral clocks from the
   light-driven SCN clock^[134]39–[135]41. We found that diurnal
   regulation of ChP TTR protein levels was dependent on feeding. The high
   dark-phase TTR protein expression was maintained in ChP from mice
   receiving dark-phase restricted feeding but was equalized in ChP from
   mice restricted to light-phase feeding, while transcript levels were
   unaffected (Fig. [136]3d, Supplementary Fig. [137]3e–h). These results
   demonstrate that the ChP responds to external cues (e.g., nutrition or
   activity) to modulate its own output, in concert with the molecular
   clock. These results both suggest that the ChP and CSF systems
   functionally respond to complex nutritional and behavioral input and
   provide a framework for testing roles of specific inputs into the
   system.

Tracking ChP and TTR rhythmicity ex vivo in ChP explants

   Because ChP translation changed diurnally, we further refined the
   timeline of ChP output through the day. We used ChP explants^[138]42
   from Per2:luc mice in which luciferase bioluminescence luminometry
   reports the expression of the circadian clock gene, Period2
   (Per2)^[139]43 (Fig. [140]3e). We tracked ex vivo ChP rhythmicity over
   14 days, and from these readouts, calculated the average period length
   to be just under 24 h (Fig. [141]3e–g). This period length was
   indistinguishable between endogenous explants and after addition of the
   potent zeitgeber dexamethasone (Fig. [142]3f, g), which, consistent
   with its endogenous ability to synchronize clocks^[143]44,
   resynchronized the tissue without altering the period length. In
   parallel, we found significant circadian oscillations in key clock
   components Per2 and Bmal1 by RT-qPCR (Supplementary Fig. [144]1,
   [145]S3a). These results corroborate previous studies showing that ChP
   maintains endogenous rhythmicity^[146]14,[147]17, and add the
   observation that ChP tissue is responsive to extrinsic glucocorticoid
   cues.

   To test for circadian expression of TTR within individual animals
   rather than from cumulative terminal samples collected at each time, we
   used genome editing to generate a Ttr^mNeonGreen reporter mouse where
   TTR harbors a C-terminal mNeonGreen tag (Fig. [148]3h, Supplementary
   Figs. [149]3d, [150]4a). The protein appropriately localized to ChP
   epithelial cells (Fig. [151]3h, Supplementary Fig. [152]3d), and was
   secreted into CSF (Supplementary Fig. [153]4b). Ex vivo 3-day imaging
   of ChP explants generated from Ttr^mNeonGreen mice indicated that TTR
   protein levels in ChP tissue cycle across circadian time (Fig. [154]3i,
   j). We validated these findings by immunoblotting lysates of ChP
   collected at 3 h intervals. TTR protein was substantially upregulated
   in ChP at 9 p.m., 2 h after lights-off, and remained relatively high
   throughout the dark phase (Fig. [155]3k), as in liver (Supplementary
   Fig. [156]3j). Using multi-day ex vivo monitoring, we confirmed prior
   reports of ChP circadian rhythmicity ex vivo (Myung et al., 2018), and
   then developed an imaging pipeline to track endogenous patterns of
   ChP-TTR expression in real-time. Collectively, we demonstrate that
   ChP-TTR diurnal protein level variations are post-transcriptionally
   regulated and suggest that ChP-TTR expression peaks around lights-off.

   We then asked how ChP-TTR affected CSF-TTR and found that after the
   onset of the dark phase, CSF-TTR levels (relative to the stably
   expressed iron binding protein Transferrin (Tf)), increased with a
   slight delay such that by 10 p.m., CSF-TTR was increased compared to
   9 p.m., and CSF-TTR remained high at 11 p.m.^[157]45–[158]47
   (Fig. [159]3l). Because a key role of TTR is to transport thyroid
   hormone, we hypothesized that levels of CSF thyroid hormone would
   correspond with changing TTR availability. To test this idea, we
   developed a liquid chromatography-mass spectrometry (LC-MS) method for
   CSF thyroid hormone detection, where we optimized several parameters,
   including chromatography-based analyte separation, ionization,
   detection, and chemical preservation. Focusing on the time during which
   the TTR transition occurs, between 6 pm and midnight (Fig. [160]3k), we
   found that increased CSF-TTR was accompanied by concurrent changes in
   multiple polar metabolites in CSF including an increase in active
   thyroid hormone T3 (triiodothyronine) (Fig. [161]3m) at 9 p.m., but not
   a significant change in the circulating form T4 (thyroxine)
   (Supplementary Fig. [162]3k, l). Collectively, application of several
   analysis modalities revealed that global changes in ChP protein
   expression track in register with CSF-TTR levels. Consistent with a
   role for CSF-TTR in thyroid hormone transport and retention, CSF-T3
   increased during the dark phase when CSF-TTR was higher. Because
   thyroid hormone levels play critical roles in neurodevelopment and
   metabolism as well as in depression, mania and other psychiatric
   manifestations^[163]48–[164]52, these data may have implications for
   thyroid sensitive brain functions and pathology.

Real-time tracking of CSF-TTR levels throughout the day in awake mice

   We sought to determine whether the time course of changes in ChP TTR
   resulted in parallel changes in TTR secretion into the CSF. Serial CSF
   sampling for immunoblotting is not suitable for this experiment because
   as a terminal procedure, it introduces both inter-individual
   variability and stress-related handling artifacts which may impair
   measurement accuracy. We overcame these challenges by adapting fiber
   photometry to the CSF in the Ttr^mNeonGreen mouse line. Adult wild-type
   and Ttr^mNeonGreen mice were outfitted with bilateral LV cannulae and
   optical fibers for freely-moving photometry recordings from CSF lasting
   between 24 and 96 h (Fig. [165]4a, b, Supplementary Fig. [166]4c–f).
   CSF mNeonGreen fluorescence showed a robust and sustained increase
   across the first lights-off transition, with the rise in CSF-TTR signal
   beginning up to two hours before the lights were extinguished
   (Fig. [167]4c). More than half of the Ttr^mNeonGreen photometry
   recordings demonstrated peak response magnitudes (defined as the
   average signal between 8:30 p.m. and 9:30 p.m.) above the range of
   autofluorescence levels observed in wild-type mice. Among these
   recordings showing robust increases in TTR levels, 20% of the peak
   signal was achieved a median of 66.9 min before lights-off
   (Fig. [168]4d–f, Supplementary Fig. [169]4g). The real-time readout of
   CSF fluorescence enabled by the photometry method thus demonstrates
   that the diurnal rhythm of CSF-TTR anticipates the light transition
   rather than being a direct consequence of light-sensitive brain
   activity. This result suggests that CSF-TTR is likely regulated by an
   intrinsic timing mechanism. Notably, a smaller subset of mice
   demonstrated a nearly flat signal within the range of wild-type
   variability, and several outliers showed a slow decrease in mNeonGreen
   signal over the same timescale, likely reflecting inter-individual
   differences in cannula placement as well as diurnal rhythm disruptions
   due to handling stress.

Fig. 4. CSF TTR is dynamic across the day.

   [170]Fig. 4
   [171]Open in a new tab

   a Schematic of the freely-moving photometry setup, consisting of a
   470 nm LED light source, CMOS camera for signal collection, and patch
   cord to the animal. Inset shows the geometry of the optical fiber in
   the cannula relative to the ventricular system of the mouse. b Overview
   of all summarized photometry recordings, with each bar indicating the
   start (on the left) and stop times (three animals recorded
   simultaneously in each recording). The first 8 h (light-dark
   transition) of both the light–dark and dark-dark recordings were
   combined for analyses and subsequently treated separately for the later
   periods when the light patterns were distinct. Two Ttr^mNeonGreen and
   one wild-type animal were excluded from analysis for health reasons. c
   Summary of all Ttr^mNeonGreen and wild-type recordings in the first 8-h
   period around the first lights-off. N = 25 biologically independent
   Ttr^mNeonGreen animals across seven independent experiments and N = 5
   biologically independent wild-type animals across two experiments. Data
   are presented as mean values ± standard error of the mean (SEM). d Peak
   responses, defined as the average signal over the 8:30 p.m.—9:30 p.m.
   interval, of all Ttr^mNeonGreen and wild-type recordings summarized in
   c. Horizontal lines are drawn at the minimal and maximal wild-type
   signals, and Ttr^mNeonGreen recordings are categorized based on these
   subdivisions into strong responders (top group, red), wild-type-like
   responders (middle group, peach), and dropping signals (bottom group,
   blue). e Heatmap showing each individual recording summarized in c,
   with subgroups color-coded as in d. Green tick marks on the heatmap
   indicate the times where each strongly responding trace reaches 20% of
   its peak signal (average between 8:30 p.m. and 9:30 p.m.). f Summary of
   the green tick marks in the heatmap in e, indicating that 20% peak
   response is reached a median of 66.9 min before lights-off. Box plots
   in d, f show median value at the red central bar, with the bottom and
   top edges of the box indicating the 25th and 75th percentiles,
   respectively. The whiskers extend to the most extreme values not
   considered outliers, and outliers (more than 1.5 times the
   interquartile range away from median) are plotted individually and
   marked with a red ‘+’ symbol. Male mice were analyzed. Source data are
   provided as a Source Data file (Source Data).

   Next, we continuously monitored mNeonGreen fluorescence across several
   days. We found that the trend of increasing CSF fluorescence in the
   night and decreasing fluorescence in the day persisted both during
   normal light–dark cycling as well as during complete darkness with a
   period just under 24 h (Fig. [172]4b, Supplementary Fig. [173]4h–j).
   Notably, despite a decay of the mNeonGreen signal over tens of hours
   due to photobleaching, the modulation amplitude over a 24-h cycle
   remained significantly greater in Ttr^mNeonGreen mice relative to
   wild-type controls a day or more into the recording, indicating that
   the signal remained above background autofluorescence (Supplementary
   Fig. [174]4k). Collectively, these results suggest that the diurnal
   rhythm of CSF TTR is not strictly determined by the light cycle, but
   the pattern of expression is maintained even in the free-running state,
   when the light cues disappear, suggesting a level of autonomy from the
   light-driven clock cues.

ChP metabolism is diurnally regulated

   In addition to secreted proteins, mRNA transcripts encoding ribosome
   components, mitoribosome subunits, and components of the electron
   transport chain oxidative phosphorylation pathway were preferentially
   associated with ribosomes during the dark phase (Figs. [175]2b, f,
   [176]5a–c). Since many components involved in aerobic respiration were
   upregulated at 9 p.m., we hypothesized that ChP capacity for oxygen
   consumption also differed across times of day. We used Agilent Seahorse
   XFe technology to monitor oxygen consumption as an index of the
   metabolic status using LV ChP explants taken from mice at 9 a.m. and
   9 p.m. (Supplementary Fig. [177]5c–e). While we identified large
   changes in metabolic components from serial samples of acutely
   collected ChP (Fig. [178]5b, h, i), we did not observe functional
   metabolic differences in oxygen consumption or ATP production between
   9 a.m. and 9 p.m. in ChP explants in constant 0.18% glucose
   (Supplementary Fig. [179]5d, e). For reference, normal mouse blood
   glucose is 80–100 mg/dL (0.08–0.1%) between fasting and feeding, and
   normal CSF glucose is usually ~60% of the plasma level^[180]53. We
   presume that explanted ChP tissue rapidly adapted to the glucose-rich
   media of the assay, precluding accurate testing of diurnal ChP
   respiration ex vivo.

Fig. 5. ChP metabolic components and metabolites are diurnally regulated.

   [181]Fig. 5
   [182]Open in a new tab

   a–c Schematics of the mitochondrial transport, the citric acid (TCA)
   cycle, and electron transport chain components. Those components that
   show altered enrichment on the ribosome at either 9 a.m. (orange) or
   9 p.m. (blue). Listed components include those that are significantly
   associated with ribosomes at either 9 a.m. or 9 p.m. and those that are
   enriched on ribosomes vs. off ribosomes at either 9 a.m. or 9 p.m. d
   Heatmap of top 25 changed metabolites in 9 a.m. vs 9 p.m. ChP. e
   Metabolite log[2] (ratio) for TCA intermediates in 9 a.m. and 9 p.m. LV
   ChP. N = 8 biologically independent animals at each time in one
   experiment. f Metabolite log[2] (ratio) for pentose phosphate pathway
   (PPP) intermediates in 9 a.m. (orange) and 9 p.m. (blue) LV ChP. N = 8
   biologically independent animals at each time in one experiment. g
   Values of common redox electron donor and recipient pairs followed by
   the log[2] (reduced: oxidized ratio) showing increased oxidation at
   9 p.m. (blue) in LV ChP. N = 8 biologically independent animals at each
   time in one experiment; Student’s two-tailed unpaired t test. Data are
   presented as mean values ± standard deviation (SD). h Immunoblotting
   for citrate synthase (CS), core components of the electron transport
   chain (OXPHOS; CV-Atp5a; CIII-Uqccrc2; CIV-Mtco1; CII-Sdhb; C1-Ndufb8),
   mitofusin2 (Mtfn2) showed enriched mitochondrial components during the
   9 p.m. dark phase. i Quantification of citrate synthase (CS) intensity
   in immunohistochemistry of LV ChP epithelial cells showed a shift
   toward higher intensity expression of CS at 9 p.m. (blue) than at
   9 a.m. (orange). ****p < 0.0001; Kolmogorov–Smirnov test, N = 5
   biologically independent animals at each time across 2 independent
   experiments. Male mice were analyzed. Source data are provided as a
   Source Data file (Source Data).

   We overcame these ex vivo technical hurdles by adapting LC-MS targeted
   metabolomics that interrogated 250 small molecules covering major
   central carbon metabolism pathways, which allowed us to investigate in
   vivo changes in ChP metabolism. ChP tissues were collected at 9 a.m.
   and 9 p.m. and immediately processed and analyzed on our metabolomics
   platform, which allowed for a more precise and unbiased view of the
   tissue’s metabolome. Independent clustering reproducibly segregated the
   two populations of tissue and identified significantly different
   metabolites at these two times (Fig. [183]5d, Supplementary
   Fig. [184]5f, g). Consistent with differentially expressed citric acid
   cycle (TCA) genes (Fig. [185]5b), metabolites of the TCA reflecting
   oxidative phosphorylation through the electron transport chain (ETC)
   were increased during the dark phase (Fig. [186]5e). By contrast,
   metabolites indicating shuttling of glucose to the pentose phosphate
   pathway (PPP) accumulated in ChP during the light phase (Fig. [187]5f).
   This major metabolic shift indicates that local ChP metabolism is an
   output of diurnal cues (intrinsic or extrinsic). Strikingly, redox
   metabolites in the ChP, including NADH, NADPH, and GSH, were
   preferentially detected in their oxidized state during the dark phase,
   with accumulation of products of the ETC including ATP (Fig. [188]5g,
   Supplementary Fig. [189]5h) and a consistent decrease in the
   reduced:oxidized ratio of major redox metabolites including
   NADPH:NADP^+, GSH:GSSG and NADH:NAD^+ (Fig. [190]5g). The increased ATP
   and oxidated species are consistent with the increased mTOR-effector
   signaling observed during the dark phase as mTORC activation mediates
   oxidative phosphorylative metabolism^[191]54 (Fig. [192]1a). In
   addition to reflecting diurnal oxidative shifts, this newly generated
   differential ChP metabolite dataset is hypothesis-generating. For
   example, significant diurnal shifts in 2-hydroxyglutarate levels could
   suggest circadian differences in important physiological processes
   including responses to hypoxia and chromatin modifications^[193]55
   (Supplementary Fig. [194]5i).

   Consistent with higher levels of oxidative metabolism during the
   dark phase and with ChP-TTR levels depending on metabolic cues like
   feeding, the citric acid cycle enzyme Citrate synthase (CS), oxidative
   phosphorylation (OXPHOS) components (CV-ATP5A; CIII-UQCCRC2; CIV-MTCO1;
   CII-SDHB; C1-NDUFB8), and Mitofusin2 (MFN2) were all upregulated at
   9 p.m. in LV ChP (Fig. [195]5h). In addition, the amount of CS within
   individual ChP epithelial cells was increased (Fig. [196]5i), although
   the number of mitochondria remained unchanged in each ChP epithelial
   cell (Supplementary Fig. [197]5a, b). Collectively, our data suggest
   that the ChP maintains an endogenous circadian rhythm that 1) diurnally
   regulates metabolic processes including TCA, ETC, and overall oxidative
   state, and 2) adapts on a relatively fast timescale in response to
   external systemic nutritional cues.

ChP barrier components and microstructure are diurnally regulated

   ChP epithelial cells comprise the blood-CSF barrier, a key brain
   barrier regulating selective access of systemic cues to the
   CNS^[198]56. Because barrier components were differentially regulated
   in our diurnal TRAP data, we next tested whether the blood-CSF barrier
   properties are diurnally regulated. Extracellular matrix (ECM) and
   adhesion proteins are necessary for appropriate circadian rhythmicity
   in other epithelia^[199]57–[200]60 and in the BBB^[201]61. Pathway
   enrichment analysis (Advaita) of our data revealed endocytosis
   (p = 0.004, FDR correction), and gene ontology (GO) enrichment analysis
   identified vesicle-mediated transport (p = 9.97 × 10^−5, FDR
   correction) to be significantly upregulated in ChP during the day
   (Figs. [202]2h, [203]6a, [204]S6a), consistent with a more permeable
   blood-CSF barrier during the light phase. Barrier components showed
   TRAP profiles consistent with higher levels of translation during the
   light phase (Fig. [205]6a). While four of these candidates demonstrated
   mild 24 h rhythmicity at the transcript level (Fig. [206]6b) [integrin
   beta-8 (Itgb8), solute carrier family 7 member 8 (Slc7a8),
   cadherin3/p-Cadherin (Cdh3), ATP binding cassette subfamily F member 3
   (Abcf3), and intracellular trafficking component sorting nextin 12
   (Snx12)], transcriptional regulation alone is insufficient to explain
   the drastic variation in their ribosomal loading between 9 a.m. and
   9 p.m. In fact, Cdh3 was the only one of these genes to have a peak in
   transcript level in the middle of the day. The other genes showed a
   striking phase reversal in that transcript levels peaked around 9 p.m.
   while ribosomal loading was clearly higher at 9 a.m. Notably, Chmp1b
   mRNA was not rhythmic even though TRAP revealed differential ribosome
   association of its transcripts. These data suggest that translational
   circadian regulation can be significantly out of phase from transcript
   rhythmicity, an important consideration in circadian studies relying on
   transcriptional readouts (Fig. [207]6b, Supplementary Fig. [208]6a).
   Diurnally regulated solute carriers included neutral amino acid
   transporter Slc7a8 (Lat2) known to maintain minimal CSF amino acid
   concentrations^[209]62 and zinc transporters (Slc39a9, Slc30a5,
   Slc30a1) necessary for function of other epithelia^[210]63
   (Fig. [211]6a, b, [212]S6a). These data support a potential
   permeability, absorption, or clearance role associated with the late
   afternoon peak of these family members in ChP (Fig. [213]6a, b).

Fig. 6. ChP barrier components and permeability are diurnally regulated.

   [214]Fig. 6
   [215]Open in a new tab

   a TRAP candidates associated with barrier function. FPKM for each
   individual mouse is shown at 9 a.m. (orange circles) and 9 p.m. (blue
   squares) for those transcripts associated with RPL10A (solid) and those
   in the supernatant (open). N = 3 biologically independent samples (LV
   ChP pooled from 3 animals per sample). Data are presented as mean
   values ± standard deviation (SD). b RT-qPCR analysis of the expression
   of barrier components Itgb8, Slca8, Chd3, Abcf3, Chmp1b, and Snx12 in
   LV ChP every 3 h. Itgb8, Slca8, Chd3, Abcf3, and Snx12 showed
   significant rhythmicity by RAIN analysis. N = 4 biologically
   independent animals at each time across 2 independent experiments. c
   Experimental design for barrier permeability assay using intravenous
   (i.v) injection of horseradish peroxidase (HRP) followed by
   transmission electron microscopy. d Example of HRP-filled vesicles
   (green circles) taken up during the 7-min incubation period within the
   peri-junctional area (yellow shading). Vesicles outside of the
   quantified region are circled in brown. e Quantification of total
   number of HRP-filled vesicles in the peri-junctional area. Each point
   represents one junction, data combined from three animals at each
   timepoint. Welch’s two-tailed unpaired t test, N = 3 biologically
   independent animals per time in one experiment. f Example of the
   apical-basal distances calculated for HRP-filled vesicles in the
   peri-junctional area. Quantification of the (apical-basal)/apical ratio
   (0 = basal (blue); 1 = apical (red) for HRP-filled vesicles in the
   peri-junctional area at 9 a.m. (orange) and 9 p.m. (blue).
   Kolmogorov–Smirnov test, N = 3 biologically independent animals per
   time in one experiment. g Example of tight junctions at apico-lateral
   surface of ChP epithelial cells. Dotted box indicates junction area and
   bracket indicates width. Scale bars: top = 1 μm; bottom zoom = 500 nm.
   h Quantification of distances between each cell membrane within the
   tight junction. Each point is the average of 5 distances per junction
   for 10 junctions in a single animal. *p < 0.05; Welch’s two-tailed
   unpaired t test, N = 3 biologically independent animals per time in one
   experiment. Data are presented as mean values ± standard error of the
   mean (SEM). Male mice were analyzed. Source data are provided as a
   Source Data file (Source Data).

   To test this idea, we next compared ChP epithelial barrier properties
   at 9 a.m. vs 9 p.m. by adapting a traditional blood–brain barrier
   assessment technique^[216]64 of intravascular horseradish peroxidase
   (HRP) labeling to observe ChP transcytosis and tight junction
   morphology (Fig. [217]6c). We did not observe significant differences
   in the number of transcytosed vesicles from the blood into ChP
   epithelial cells along the basolateral surface of the cell-cell
   junction (Fig. [218]6d, e, Supplementary Fig. [219]6b). While we noted
   a trend toward more vesicles localized near the apical surface during
   the light phase, there was variability among individuals (Fig. [220]6f,
   Supplementary Fig. [221]6c). The apical tight junction, however, was
   significantly and substantially wider during the 9 a.m. light phase
   (Fig. [222]6g, h, Supplementary Fig. [223]6d, e), potentially
   indicating a more permeable ChP barrier during the light phase. Taken
   together, our results show rhythmicity in structural components that
   support a model that the blood-CSF barrier, like the blood–brain
   barrier (BBB), exhibits higher permeability during the light
   phase^[224]65. This model suggests potential implications for
   differential CSF/ brain access of drugs and systemic cues throughout
   the day. This sensitive set of analyses can be broadly applied to query
   ChP barrier permeability changes in other states and circumstances.

Discussion

   We developed, curated, and adapted a set of multi-modal approaches that
   comprise an integrated toolkit for analyzing the ChP-CSF axis across
   circadian time and in response to associated environmental cues.
   Additionally, data generated from low-abundance samples of mouse ChP
   and CSF analyzed with this toolkit represent valuable benchmarks for
   how the ChP and CSF change throughout the day and respond to external
   cues. We used TRAP ribosomal pulldown, ex vivo ChP explant luminometry
   and microscopy, in vivo CSF fiber photometry using a Ttr^mNeonGreen
   reporter mouse generated for this purpose, CSF and ChP metabolomics,
   and ChP barrier analysis. Using these tools, we demonstrated that ChP
   translation and protein synthesis were regulated diurnally, with higher
   levels of translation taking place during the dark phases when these
   animals are more active. This change in translation involved distinct
   sets of proteins synthesized by the ChP during the dark vs. light
   phases. These proteins were involved in the core ChP functions of
   secretion, metabolism, and brain barrier properties. We focused on TTR
   because it emerged as a top candidate from TRAP analysis and is a key
   component of ChP and CSF. We found that TTR protein abundance responded
   to feeding cues and required the molecular clock component Bmal1.
   Ultimately, diurnal regulation of ChP-TTR expression modulated CSF-TTR
   levels. Concurrently, ChP metabolism was substantially different
   between day and night with more oxidative phosphorylation during the
   dark phase. The data were consistent with a more permeable ChP barrier
   during the light phase. Together, these tools described concerted daily
   changes in the ChP and CSF with far-reaching consequences for
   understanding the composition of this essential and surgically
   accessible fluid.

   Protein translation is emerging as a fundamental process by which
   circadian rhythmicity controls tissue functions throughout the body.
   For example, mTOR/4E-BP1-mediated translational control regulates
   entrainment and synchrony of the SCN master clock^[225]66,[226]67. In
   liver, ribosome biogenesis, translation, and protein synthesis are
   downstream of core circadian components responsive to metabolic
   cues^[227]24,[228]26. At the cellular level, BMAL1 rhythmically
   interacts with translational machinery to promote protein synthesis in
   response to mTOR signaling, thereby directly connecting circadian
   timing to the control of protein production^[229]25. Previous efforts
   to identify downstream effectors of this circadian rhythmicity on ChP
   output were somewhat inconclusive—mRNA quantification has either
   identified no change, as in the case of the water channel Aqp1^[230]17,
   or smaller changes for ApoJ and Ttr in rats^[231]38. TTR itself could
   be induced in cultured ChP cells by glucocorticoids, and upregulated in
   liver, ChP, and CSF by acute and sometimes chronic stress^[232]68. The
   bursts in ChP translation that we observed in the dark phase are likely
   followed by protein clearance (degradation or removal via CSF outflow)
   to manifest the quick shifts in protein availability. Consistent with
   this hypothesis, pathway analysis suggests increased ChP ubiquitin
   mediated proteolysis during the light phase (p = 0.002, FDR correction)
   and previous studies indicate that CSF production and distribution
   shift diurnally^[233]2,[234]5. Our finding that ChP diurnal output is
   regulated post-transcriptionally suggests that nutrient sensing (e.g.,
   via mTOR) in the ChP may be an upstream regulator of ChP diurnal
   functions including CSF secretion and brain barrier permeability.

   The ChP supplies most CSF-TTR^[235]37 and altered CSF-TTR has been
   implicated in neurologic diseases. For example, low CSF-TTR is
   associated with depression^[236]52,[237]69,[238]70. CSF-TTR may be
   protective against Aβ toxicity as a result of its ability to bind
   Aβ^[239]71. Indeed, Aβ clearance changes during sleep vs. waking
   states^[240]3. TTR is associated with oxidative stress in other
   systems^[241]72, potentially linking the metabolic oxidative shifts
   with dark phase TTR availability. TTR is also the major transporter of
   thyroid hormone from the ChP epithelial cells into the
   CSF^[242]36,[243]37,[244]73,[245]74 and can prevent loss of thyroid
   hormone from the CSF into the bloodstream in
   hypothyroid animals^[246]75. Thyroid hormone levels cycle in a
   circadian manner in serum^[247]76,[248]77 and in brain tissue^[249]78,
   but previously published data are not conclusive on the origin of
   thyroid hormone cycling in the brain^[250]51. Our data indicate that
   varying ChP-TTR levels modulate thyroid hormone (T3) availability in
   CSF. The rise in CSF T3 corresponds with higher serum T4, suggesting
   that dark phase ChP TTR could be upregulated to transport systemic
   thyroid hormone into the CSF. Further supporting a role for TTR in
   thyroid hormone availability, pathway enrichment analysis (Advaita) of
   9 a.m. vs. 9 p.m. ChP TRAP data identified thyroid hormone synthesis
   and signaling (p = 0.013, FDR correction) as differentially regulated
   pathways between light and dark phases.

   The ChP displayed a concerted diurnal metabolic shift, including
   increased oxidation signal during the dark phase. Significantly
   different metabolite profiles were associated with both ChP and CSF
   between 9 a.m. and 9 p.m., with others likely to have nadirs and
   zeniths in accumulation that were not captured by these two timepoints.
   However, in explants (Supplementary Fig. [251]5), we presume that the
   Seahorse media, which contains high glucose, overrode the intrinsic
   metabolic activity of ChP tissues, especially since we found that at
   least some ChP circadian properties are feeding dependent
   (Fig. [252]3d). This suggests that care should be exercised in explant
   preparations and is why subsequent ex vivo analyses were performed on
   purposefully synchronized tissue and monitored for multiple days.
   Metabolic responses to diurnal cues are common in other tissues as
   well, with the circadian clock controlling metabolism at the level of
   transcription and translation^[253]20,[254]79. In liver, protein
   expression of Krebs cycle and oxidative respiratory chain enzymes
   oscillates, and this rhythmicity is blunted in Per1/2-null and
   Bmal1-null mice^[255]33,[256]80. Metabolic output is also dependent on
   circadian rhythm in other cell types including cardiomyocytes^[257]81,
   skeletal muscle^[258]82, hippocampus^[259]83, pancreatic
   β-cells^[260]84, and macrophages^[261]85. Gut epithelial metabolism
   demonstrates circadian rhythmicity, particularly of cytochrome P450
   family members^[262]86. Consistently, our data show higher dark phase
   cytochrome P450 monooxygenases that target fatty acids/xenobiotics
   (Cyp2j6: log[2]FC = −3.047, p adj = 0.002; Cyp2u1: log[2]FC = −1.851, p
   adj = 0.009). Circadian disruptions link metabolic disease and
   cognitive decline^[263]87–[264]89. Our data showing diurnal changes in
   ChP metabolic components motivate actively integrating ChP into models
   of these and other diseases. These data corroborate a large body of
   work indicating metabolic changes across time of day in multiple
   tissues^[265]34 and place the ChP into the context of systemic
   metabolism, potentially reflecting its role as a key interface between
   systemic circulation and the CSF.

   The diurnal cycles in cell adhesion components, solute transporters,
   efflux pumps, and endocytosis that we observed likely modulate barrier
   properties and could have important clinical implications for
   chronopharmacology—drug treatment that takes the body’s circadian
   rhythm into consideration. In fact, 85% of trials for drugs with
   half-lives under 15 h showed dosing time dependence compared with just
   39% for longer-acting drugs^[266]90,[267]91. Some diseases, including
   psychiatric symptoms, manifest morbidities at specific circadian
   times^[268]91. Consistent with our observed ChP permeability changes,
   BBB permeability including carrier-mediated transport, cell-cell
   junction permeability, and efflux pumps also varies diurnally, with a
   more permeable barrier during light phases^[269]56,[270]65. Recently,
   neuronal activity (higher during waking hours) was shown to activate
   endothelial cell ABC efflux transporter expression at the BBB through a
   Bmal1-dependent mechanism^[271]61. Similarly, intestinal barrier
   properties change in response to feeding and other circadian cues as is
   seen in expression of tight junction proteins Occludin and
   Claudin-1^[272]92. Because the ChP is a key brain barrier that can
   modulate drug efflux^[273]56, our data showing diurnal differences in
   ChP barrier properties have implications for waste clearance, immune
   cell trafficking, and CNS drug efflux/ influx.

   Abnormal CSF components have been associated with a number of
   neurologic conditions that co-present with circadian disruptions,
   including hydrocephalus, autism spectrum disorder, Alzheimer’s disease,
   bipolar disorder, and schizophrenia^[274]93–[275]99. For example, sleep
   disruption is a diagnostic criterion for major depression, bipolar
   disorder, post-traumatic stress disorder, generalized anxiety, and
   other mood disorders^[276]100. While CSF components, including those
   that are diurnally regulated, can originate from outside the ChP (e.g.,
   peripheral adrenal glands (cortisol), hypothalamus (corticotropin
   releasing hormone, CRH), SCN (vasoactive intestinal peptide (VIP),
   arginine vasopressin (AVP), gastrin-releasing peptide (GRP)), or pineal
   gland (e.g., melatonin)^[277]10,[278]12,[279]101–[280]104, this study
   reveals large-scale diurnal changes in ChP translation that lead to
   altered CSF contents. Additionally, metabolism and mitochondrial
   function are disrupted in individuals also at high risk for
   psychosis^[281]105,[282]106 and bioenergetics profiles are altered
   during late onset Alzheimer’s disease^[283]107. The current study
   defines a proteostatic and metabolic framework in the ChP that changes
   diurnally, and in the case of TTR, in response to systemic cues like
   feeding. These data open avenues for future studies to identify whether
   such metabolic and CSF composition changes are causes or symptoms of
   neurological disease. Specifically, our results showing the ability of
   behavioral interventions to alter the CNS fluid environment represent
   an initial step to identifying daily CSF variation and suggest methods
   of controlling its composition.

   To generate these tools, we overcame several challenges and encountered
   new limitations. (1) To enrich for actively translated transcripts, we
   used the TRAP ribosomal pulldown technique. However, TRAP has a
   preference for capturing changes in metabolic translation, as ribosomes
   containing RPL10A preferentially translate metabolic, cell cycle, and
   developmental targets^[284]30. Transcripts that are preferentially
   translated by other ribosomal subunits are not part of this analysis.
   (2) Circadian transitions are continuous, but we performed the initial
   discovery stages using TRAP ribosome pulldown at only 2 orthogonal
   timepoints, and therefore we cannot rule out the possibility that some
   important circadian ChP protein changes were not discovered using this
   method. (3) This study emphasized changes in protein production, but
   quick circadian shifts in protein availability will also require
   analysis of protein degradation and turnover.

Methods

   This research complies with all relevant ethical regulations and was
   approved by Boston Children’s Hospital IACUC and Boston Children’s
   Hospital Institutional Biosafety Committee.

Mice

   The Boston Children’s Hospital IACUC approved all experiments involving
   mice in this study. Adult CD1 male mice were obtained from Charles
   River Laboratories. Mice with germline Tg(FOXJ1-cre)F26Htzm (referred
   as Foxj1-cre) were imported from Washington University^[285]31 and bred
   in-house. Transgenic floxed EGFP-L10a mice
   [B6;129S4-Gt(ROSA)26Sor^tm9(EGFP/Rpl10a)Amc/J], Bmal1-null mice
   [B6.129-Arntl^tm1Bra/J]^[286]40 and Per2-Luciferase mice
   [B6.129S6-Per2^tm1Jt/J] were purchased from Jackson Laboratories (Stock
   Numbers: 024750, 009100 and 006852, respectively). The Ttr^mNeonGreen
   mouse line was generated as detailed below. Animals were housed in a
   temperature- and humidity-controlled room (70 ± 3 °F, 35–70% humidity)
   on a 12 h light/12 h dark cycle (7 a.m. on 7 p.m. off) and had free
   access to food and water, unless otherwise specified. All mice younger
   than postnatal day 10 were allocated into groups based solely on the
   gestational age without respect to sex (both males and females were
   included). For studies involving mice older than 10 days, only male
   mice were included.

TRAP

   Foxj1:Cre x floxed EGFP-L10a transgenic lines were crossed to generate
   mice heterozygous for each transgene Foxj1:Cre^+/−; floxed
   EGFP-L10a^+/−. Mice were kept in standard housing with 12 h light/ 12 h
   dark cycle (7 a.m. on/ 7 p.m. off), fed ad libitum, and aged 8 weeks.
   For TRAP, fresh brain tissue was dissected (N = 3 at each time — 9 a.m.
   and 9 p.m., each N included LV ChP pooled from 3 mice). Whole ChP from
   the lateral ventricle was harvested using #5 forceps and a scalpel in
   1× HBSS. To collect the LV ChP, the cerebellum was separated from the
   mid- and forebrain using a scalpel, followed by a bilateral cut along
   the midline to separate the brain into two hemispheres. Each hemisphere
   was stabilized with forceps and a third of the rostral end was cut off.
   The medial cortex was removed with a scalpel to expose the ventricle
   and free margin of the ChP. The attached LV ChP was gently separated
   from the hippocampus/fornix using forceps and immediately extracted for
   TRAP RNA purifications^[287]28. RNA quality was assessed using
   Bioanalyzer Pico Chips (Agilent, 5067-1513) and quantified using
   Quant-iT RiboGreen RNA assay kit (Thermo Fisher Scientific
   [288]R11490). Libraries were prepared using Clonetech SMARTer Pico with
   ribodepletion and Illumina HiSeq to 50NT single-end reads. Sequencing
   was performed at the MIT BioMicroCenter.

Sequencing data analysis

   The raw fastq data of 50-bp single-end sequencing reads were aligned to
   the mouse mm10 reference genome using STAR 2.4.0 RNA-Seq
   aligner^[289]108. The mapped reads were processed by htseq-count of
   HTSeq software^[290]109 with mm10 gene annotation to count the number
   of reads mapped to each gene. The Cuffquant module of the Cufflinks
   software^[291]110 was used to calculate gene FPKM (Fragments Per
   Kilobase of transcript per Million mapped reads) values. Gene
   differential expression test between transcripts associated with RPL10a
   and transcripts not associated with RPL10a was performed using DESeq2
   package^[292]111 and differentially expressed genes were defined as
   DeSeq2 adjusted p < 0.05; |log[2] FC| > 0.3. To cast a wider net for
   gene differential test between animal groups at different times,
   differentially expressed genes identification was performed using
   CuffDiff module of Cufflinks software^[293]110 and differentially
   expressed genes were defined as CuffDiff; adjusted p < 0.05; |log[2]
   FC| > 0.4. All tests were done with the assumption of negative binomial
   distribution for RNA-Seq data. All analyses were performed using genes
   with FPKM > 1, which we considered as the threshold of expression
   (Figs. [294]1, [295]2 source data). Hierarchical clustering of z-scores
   reveals consistency among all three samples (Fig. [296]2).

Sequencing pathway and motif analysis

   Functional annotation clustering was performed using DAVID
   v6.7^[297]112. Gene ontology (GO) analysis was performed using
   AdvaitaBio iPathway guide V.v1702. Enrichment vs. perturbation analysis
   was performed by AdvaitaBio iPathway guide V.v1702 and allows
   comparison of pathway output perturbation and cumulative gene-set
   expression changes. In brief, the enrichment analysis is a
   straightforward gene-set enrichment over representation analysis (ORA)
   considering the number of differentially expressed genes (DEGs) that
   are assigned to a given pathway. The enrichment value is expressed as a
   proportion of enriched members to total genes in a defined pathway and
   a p-value (Fisher) is calculated for this score^[298]113. Motif
   analyses were performed using SignalP (v5.0)^[299]114,[300]115 and
   TMHMM (v2.0)^[301]116. Predicted secreted gene products were validated
   by Human Protein Atlas and UniProtKB/Swiss-Prot.

Seahorse metabolic analysis

   ChP explants were dissected in HBSS (Fisher, SH30031FS) and maintained
   on wet ice until plated. Only the posterior leaflet of the LV ChP was
   retained for analysis due to empirically determined limitations of the
   oxygen availability in the XFe96 Agilent Seahorse system. Tissue
   explants were plated on Seahorse XFe96 spheroid microplates (Agilent,
   102905-100) coated with Cell TAK (Corning), in Seahorse XF Base Medium
   (Agilent, 102353–100) supplemented with 0.18% glucose, 1mM L-glutamine,
   and 1 mM pyruvate at pH7.4 and incubated for 1 h at 37 °C in a
   non-CO[2]incubator. Extracellular acidification rates (ECAR) and oxygen
   consumption rates (OCR) were measured via the Cell Mito Stress Test
   (Agilent, 103015-100) with a Seahorse XFe96Analyzer (Agilent) following
   the manufacturer’s protocols. Data were processed using Wave software
   (Agilent). ATP production was calculated as the difference in OCR
   measurements before and after oligomycin injection, as described by the
   manufacturer’s protocol (Agilent, 103015-100). Calcein AM was used to
   normalize between wells (Invitrogen L-3224).

Sample preparation for LC-MS analysis of polar metabolites from ChP

   Mice (CD1 males) kept in circadian cabinet housing with 12-h light
   cycle (7 a.m. on/ 7 p.m. off) aged 8 weeks (N = 16 at each time—9 a.m.
   and 9 p.m.), were decapitated and brain tissue was immediately
   dissected and frozen on dry ice. ChP were extracted by brief sonication
   in 200 µl of extraction solvent (80% LC/MS-grade methanol, 20% 25 mM
   Ammonium Acetate and 2.5 mM Na-Ascorbate prepared in LC/MS water and
   supplemented with isotopically labeled internal standards (17 amino
   acids and isotopically labeled reduced glutathione, Cambridge Isotope
   Laboratories, MSK-A2-1.2 and CNLM-6245-10). After centrifugation for
   10 min at maximum speed on a benchtop centrifuge (Eppendorf) the
   cleared supernatant was dried using a nitrogen dryer and reconstituted
   in 20 µl water (supplemented with QReSS, Cambridge Isotope
   Laboratories, MSK-QRESS-KIT) by brief vortexing. Extracted metabolites
   were spun again and cleared supernatant was transferred to LC-MS micro
   vials. A small amount of each sample was pooled and serially diluted 3-
   and 10-fold to be used as quality controls throughout the run of each
   batch.

Sample preparation for LC-MS analysis of thyroid hormone metabolites from CSF
and plasma

   CSF was collected from adult (3 months old) wild-type CD1 mice. Samples
   were placed on wet ice, then spun 1000 × g for 10 min at 4 °C. The
   supernatant was collected. Per condition, 5–10 μL of fresh, cleared CSF
   was extracted in 4:6:3 chloroform:methanol:water mixture supplemented
   with isotopically labeled T3 and T4 (Cambridge Isotope Laboratories,
   CLM-7185-C and CLM-8931-PK) as well as isotopically labeled 17 amino
   acids and isotopically labeled reduced glutathione (Cambridge Isotope
   Laboratories, MSK-A2-1.2 and CNLM-6245-10). After centrifugation for
   10 min at maximum speed on a benchtop centrifuge (Eppendorf) the top,
   hydrophilic layer was transferred to a new tube, dried using a nitrogen
   dryer and reconstituted in 20 µl water (supplemented with QReSS,
   Cambridge Isotope Laboratories, MSK-QRESS-KIT) by brief vortexing.
   Extracted metabolites were spun again and cleared supernatant was
   transferred to LC-MS micro vials. A small amount of each sample was
   pooled and serially diluted 3- and 10-fold to be used as quality
   controls throughout the run of each batch.

Chromatographic conditions for LC/MS

ZIC-pHILIC chromatography for polar metabolites

   1–2 µl of reconstituted sample was injected into a ZIC-pHILIC
   150 × 2.1 mm (5 µm particle size) column (EMD Millipore) operated on a
   Vanquish™ Flex UHPLC Systems (Thermo Fisher Scientific, San Jose, CA).
   Chromatographic separation was achieved using the following conditions:
   buffer A was acetonitrile; buffer B was 20 mM ammonium carbonate, 0.1%
   ammonium hydroxide. Gradient conditions were as follows: linear
   gradient from 20 to 80% B; 20–20.5 min: from 80 to 20% B; 20.5–28 min:
   hold at 20% B. The column oven and autosampler tray were held at 25 °C
   and 4 °C, respectively.

C18 chromatography for T3/T4

   5–7 µl of reconstituted sample was injected onto an Ascentis Express
   C18 HPLC column (2.7 μm × 15 cm × 2.1 mm; Sigma Aldrich). The column
   oven and autosampler tray were held at 30 °C and 4 °C, respectively.
   The following conditions were used to achieve chromatographic
   separation: buffer A was 0.1% formic acid; buffer B was acetonitrile
   with 0.1% formic acid. The chromatographic gradient was run at a flow
   rate of 0.250 ml min^−1 as follows: 0–5 min: gradient was held at 5% B;
   2–12.1 min: linear gradient of 5 to 95% B; 12.1–17.0 min: 95% B;
   17.1–21.0 min: gradient was returned to 5% B.

MS data acquisition conditions for targeted analysis of polar metabolites and
thyroid hormones

   MS data acquisition was performed using a QExactive benchtop orbitrap
   mass spectrometer equipped with an Ion Max source and a HESI II probe
   (Thermo Fisher Scientific, San Jose, CA) and was performed in positive
   and negative ionization mode in a range of m/z = 70–1000, with the
   resolution set at 70,000, the AGC target at 1 × 10^6, and the maximum
   injection time (Max IT) at 20 msec. A narrower scan in positive mode at
   m/z = 600–800 was used for more specific detection of thyroxine
   hormones, the resolution was set at 70,000, the AGC target was
   5 × 10^5, and the max IT was 100 msec. For polar metabolites HESI
   conditions were: Sheath gas frow rate: 35; Aug gas flow rate: 8; Sweep
   gas flow rate: 1; Spray voltage: 3.5 kV (pos), 2.8 kV (neg); Capillary
   temperature: 320 °C; S-lens RF: 50; Aux gas heater temperature: 350 °C.
   For T3/T4 HESI conditions were: Sheath gas frow rate: 40; Aug gas flow
   rate: 10; Sweep gas flow rate: 0; Spray voltage: 3.5 kV (pos), 2.8 kV
   (neg); Capillary temperature: 380 °C; S-lens RF: 60; Aux gas heater
   temperature: 420 °C.

   Relative quantitation of polar metabolites was performed with
   TraceFinder 5.1 (Thermo Fisher Scientific, Waltham, MA) using a 5 ppm
   mass tolerance and referencing an in-house library of chemical
   standards. Pooled samples and fractional dilutions were prepared as
   quality controls and only those metabolites were taken for further
   analysis, for which the correlation between the dilution factor and the
   peak area was >0.95 (high-confidence metabolites) and for which the
   coefficient of variation (CV) was below 30%. Normalization for
   biological material amounts was based on the total integrated peak area
   values of high-confidence metabolites within an experimental batch
   after normalizing to the averaged factor from all mean-centered
   chromatographic peak areas of isotopically labeled amino acids and
   internal standards. For thyroid hormone peak normalization,
   isotopically labeled thyroid hormone standards were used. Where
   indicated, data was control mean-centered, otherwise data were Log
   transformed and Pareto scaled within the MetaboAnalyst-based
   statistical analysis platform (v5.0)^[302]117. All heatmap, PCA, or
   PLSDA analysis were generated using the MetaboAnalyst online platform.
   Individual one-way Anova and t-tests were performed in Prism software
   (GraphPad v7).

Tissue processing for histology

   Samples were fixed in 4% paraformaldehyde (PFA). For cryosectioning,
   samples were incubated in the following series of solutions: 10%
   sucrose, 20% sucrose, 30% sucrose, 1:1 mixture of 30% sucrose and OCT
   (overnight), and OCT (1 h). Samples were frozen in OCT.

Immunostaining

   Cryosections were blocked and permeabilized (0.3% Triton-X-100 in PBS;
   5% serum), incubated in primary antibodies overnight and secondary
   antibodies for 2 h. Sections were counterstained with Hoechst 33342
   (Invitrogen H3570, 1:10000) and mounted using Fluoromount-G
   (SouthernBiotech). The following primary antibodies were used: chicken
   anti-GFP (Abcam ab13970 (RRID: AB_300798); 1:1000), rabbit anti-pS6
   ribosomal protein (Ser 240/244) (Cell Signaling #5364 S (RRID:
   AB_10694233); 1:1000), rabbit anti-citrate synthase (cloneD7V8B) (Cell
   Signaling #14309 (RRID: AB_2665545); 1:1000), mouse anti-fibrillarin
   (Abcam ab4566 (RRID: AB_304523); 1:250). The following secondary
   antibodies were used for immunostaining: Goat anti-Rabbit IgG (H + L)
   Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 (Thermo
   Fisher A11034; 1:500); Goat anti-Mouse IgG (H + L) Highly
   Cross-Adsorbed Secondary Antibody, Alexa Fluor Plus 647 (Thermo Fisher
   A32728; 1:500); Goat anti-Chicken IgY (H + L) Secondary Antibody, Alexa
   Fluor 488 conjugate (Thermo Fisher A11039; 1:500). Citric acid antigen
   retrieval was used for fibrillarin staining (15 min steaming in 10 mM
   sodium citrate, 0.05% Tween 20, pH = 6.0). Secondary antibodies were
   selected from the Alexa series (Invitrogen/ Thermo Fisher, 1:500).
   Images were acquired using Zeiss LSM880 confocal microscope with ×20
   objective.

Quantification of nucleolar volume

   Nucleolar volume was quantified in accordance with published methods
   using Imaris in a blinded manner^[303]118–[304]121. Cryosections for
   quantification were 20 μm thick and stained with anti-fibrillarin
   antibody (Abcam, Cambridge, MA). Z-stack images were acquired using the
   ×40 objective on an LSM 700 laser scanning confocal microscope (Carl
   Zeiss, Oberkochen, Germany). Three-dimensional reconstruction was
   performed using the Surface tool in Imaris image analysis software
   version 7.7.1 (Bitplane, Zurich, Switzerland). Nucleoli with sphericity
   <0.44 or volume <0.10 μm^3 were considered staining artefacts and
   excluded. Median values of all nucleolar volumes from a single mouse
   were plotted (N = 5 mice).

OPP incorporation

   ChP OPP incorporation was performed as described by Liu et al.^[305]27.
   Mice received intraperitoneal OPP injections (50 mg/kg OPP; Life
   Technologies). One hour later, tissues were obtained and sectioned to a
   thickness of 7 mm using a cryostat. OPP signals were detected using the
   Click-iT plus OPP protein synthesis assay kits (Life Technologies)
   according to the manufacturer’s suggested procedures. Images were taken
   at 20 × (Zeiss Axio Observer D1 inverted microscope) and fluorescence
   intensity was quantified using FIJI (ImageJ). For each sample, OPP
   intensity was measured in FIJI in an ROI that included only LV ChP
   tissue.

Rhythmicity analysis of RT-qPCR data

   RAIN (Rhythmicity Analysis Incorporating Nonparametric
   methods)^[306]122 was used to analyze rhythmicity of 3-h qPCR data and
   multi-day TTR:mNeonGreen light–dark and dark-dark photometry data using
   Bioconductor version 3.10 (BiocManager 1.30.10) in R version 3.6.3
   (2020-02-029) with the period set at 24 h.

Per2- luciferase explant luminometry

   ChP explants from Per2:luc mice^[307]43 were dissected in HBSS (Fisher,
   SH30031FS) and maintained on wet ice until plated. Explants were
   transferred to 24 well plates with 500 μL of filter-sterilized
   Lumicycle media (phenol-free DMEM (Sigma D-2902), 10 mM HEPES (pH 8.0),
   4 mM sodium bicarbonate, 25 mM D-glucose, 1 × B27 (Gibco), 4mM
   L-glutamine, Penicillin/Streptomycin) freshly supplemented with 100 mM
   beetle D-luciferin (Promega, E1601) and/or 100 nM dexamethasone. PCR
   plate sealer was used to seal the wells for the duration of the
   experiment (ThermoFisher) and placed in a Lumicycle-96 (Actimetrics) in
   a water-jacketed incubator at 35˚C. Luminometry was performed by
   iterative measurement in 60 s bins, 10 times per hour. Luminometry data
   was analyzed with Lumicycle analysis software (Actimetrics). Only
   traces with a goodness-of-fit of at least 80% were included. Period was
   measured using the χ^2 function.

Immunoblotting

   Tissues were homogenized in RIPA buffer supplemented with protease and
   phosphatase inhibitors. Protein concentration was determined by BCA
   assay (Thermo Scientific 23227). Samples were denatured in 2% SDS with
   2-mercaptoethanol by heating at 37 °C (for OxPhos) or 95 °C (for TTR
   and SERPINE2) for 3 min. Samples were centrifuged at maximum speed and
   then equal amounts of proteins were loaded and separated by
   electrophoresis in a 4–15% gradient polyacrylamide gel (BioRad
   #1653320) or NuPAGE 4–12% Bis-Tris gel (Invitrogen #NP0322),
   transferred to a nitrocellulose membrane (250 mA, 1.5 h, on ice),
   blocked in filtered 5% BSA or milk in TBST, incubated with primary
   antibodies overnight at 4 °C followed by HRP conjugated secondary
   antibodies (1:5000) for 1 h, and visualized with ECL substrate. For
   phosphorylated protein analysis, the phospho-proteins were probed
   first, and then blots were stripped (Thermo Scientific 21059) and
   reprobed for total proteins. The following primary antibodies were
   used: rabbit anti-S6 ribosomal protein (Cell Signaling #2217 S (RRID:
   AB_331355); 1:1000), rabbit anti-pS6 ribosomal protein (Ser 240/244)
   (Cell Signaling #5364 S (RRID: AB_10694233); 1:1000), rabbit
   anti-citrate synthase (cloneD7V8B) (Cell Signaling #14309 (RRID:
   AB_2665545); 1:1000), rabbit-anti-mitofusin2 (clone D2D10) (Cell
   Signaling #9482 (RRID: AB_2716838); 1:1000), rabbit anti-Vinculin (Cell
   Signaling #13901 S (RRID: AB_2728768); 1:10000), Total OXPHOS Rodent WB
   Antibody Cocktail (Abcam ab110413 (RRID: AB_2629281); 1:250), rabbit
   Anti-human prealbumin (TTR) (Dako/Agilent A000202-2 (RRID: AB_578466);
   1:200), mouse anti-transferrin [3C11] (Abcam ab70826 (RRID:
   AB_1281166); 1:750), mouse anti-mNeonGreen (ChromoTek Cat# 32f6-100
   (RRID:AB_2827566); 1:1000). The following secondary antibodies were
   used for immunoblotting: Goat anti-Mouse IgG (H + L) HRP (Life
   Technologies 31430; 1:1000); Goat anti-Rabbit IgG (H + L) HRP (Life
   Technologies 31460; 1:1000). Source data are provided as a Source Data
   file (Source Data).

Quantitative RT-PCR

   For mRNA expression analyses, the ChP were collected and one quarter of
   4 V ChP or one-half of one LV ChP was frozen for RT-qPCR, while the
   rest of the tissue from an individual was frozen for protein analysis.
   RNA was isolated using the RecoverAll RNA/DNA isolation kit (Life
   Technologies A26135) for ChP and using TRIzol Reagent (Invitrogen
   15596026) for liver samples following manufacturer’s specifications for
   RNA extraction. Extracted RNA was quantified spectrophotometrically and
   50 ng was reverse-transcribed into cDNA using the Promega™ ImProm-II™
   Reverse Transcription System (Fisher PR-A3802) following manufacturer’s
   specifications with random hexamer primers (Promega PR-C1181). RT-qPCRs
   were performed in duplicate using Taqman Gene Expression Assays and
   Taqman Gene Expression Master Mix (Applied Biosystems) with Gapdh or
   Rp1p0 as an internal control. Cycling was executed using the
   StepOnePlus Real-Time PCR System (Invitrogen) and analysis of relative
   gene expression was performed using the 2^−ΔΔCT method^[308]123 and
   presented as RQ of data normalized to 9 am and the internal control.
   Technical replicates were averaged for their cycling thresholds and
   further calculations were performed with those means. The following
   mouse Taqman probes (Thermo Fisher) were used: Ttr (Mm00443267_m1),
   Serpine2 (Mm00436753_m1), Bmal1(Mm00500223_m1), Per2(Mm00478099_m1),
   Itgb8(Mm00623991_m1), Cdh3(Mm01249209_m1), Abcf3 (Mm00658695_m1),
   Slc7a8(Mm01318974_m1), Chmp1b (Mm04179599_s1), Gapdh VIC
   (Mm99999915_g1), Rp1p0 VIC (Mm00725448_s1).

Restricted feeding

   Adult mice were singly housed in cages placed in a controlled lighting
   environment away from the main mouse colony. The lights were programmed
   to come on at 7 a.m. and turn off at 7 p.m. The mice were acclimatized
   to the boxes for 5 days with ad libitum feeding. Then for 7 days, at
   each lighting change, mice were either moved between a home cage with
   water and no food, or a home cage with food and water. For control
   feeding (night feeding), mice were moved to a food-containing cage at
   7 p.m. and mice were moved to food-free cages at 7 a.m. For inverted
   feeding (day feeding), mice were moved to a food-containing cage at
   7 a.m. and mice were moved to food-free cages at 7 p.m. All cage
   changes took no more than 10 min. Mice always had access to water.

Generation of Ttr^mNeonGreen mouse line

   Ttr^mNeonGreen mice were generated on C57BL6 background using CRISPR
   technology with homology-directed repair (HDR). The template HDR
   construct was synthesized to insert the mNeonGreen sequence at the
   C-terminus of Ttr. The following gRNAs were used:

   5′-GGGGTTGCTGACGACAGCCG-3′; 5′-CCCATACTCCTACAGCACCA-3′;

   5′-AGAATTGAGAGACTCAGCCC-3′

   Generation of the mouse line was carried out by Boston Children’s
   Hospital Mouse Gene Manipulation Core. Mice were screened with 2 sets
   of PCR primers for correct insertion of mNeonGreen protein. PCR
   products were sequenced to ensure no mutations during the CRISPR
   editing process.

   PCR primer set 1:

   GAGCAAGTGACAGAGTTGCCCT; agtctcTTACTTGTACAGCTCGTCCATGCCCATC

   PCR primer set 2:

   TCTGGTGGAGGTGGATCCGGAG; GGGAAACGACGGATCGGGGAG

   The founder was bred to C57BL6 for 5 generations before experiments
   were performed. PCR primer for genotyping progeny are:

   Gentyping primer set: CAGCCATTGGGTGCCACAAC; CTGTCGTCAGCAACCCCCAGAAT.

Characterization of Ttr^mNeonGreen mice

   Immunoblotting of the ChP and CSF: The ChP was dissected, and protein
   extracted (homogenized in RIPA buffer). 50 µg total protein was loaded
   to a single well for immunoblotting. Up to 15 µl of CSF was loaded to a
   single well for immunoblotting. All CSF samples were normalized based
   on total protein level. The blots were incubated in mNeon antibody
   (Chromotek 32f6, 1:200, Mouse monoclonal) at room temperature for
   15 min before moving to 4 °C overnight. TTR and GAPDH were probed on
   the same blots after stripping.

   Immunohistochemistry: Tissue sections were stained with Hoechst and
   directly mounted. Endogenous mNeon signal was imaged by confocal
   microscopy.

   CSF fluorescence detection: CSF was collected from adult (3 months old)
   Ttr^mNeonGreen mice and their wild-type littermate controls. 10 µl CSF
   was diluted into 30 µl with PBS and loaded to a 96-well plate with
   clear bottom. Blank PBS was included as negative control. Fluorescent
   signal was measured with 488 nm excitation on a Tecan plate reader.

CSF fiber photometry recordings from freely-moving mice

   Wild-type and Ttr^mNeonGreen mice used for in vivo fiber photometry
   experiments were surgically outfitted with a titanium headpost (0.7 g,
   H.E. Parmer) and bilateral stainless steel 22XX gauge LV guide cannulae
   (Microgroup, length: 7 mm). Briefly, mice were anesthetized with 1–4%
   isoflurane in 100% O[2] and breathing rate was maintained at 1 breath
   per second. Animals were administered dexamethasone (1 mg/kg;
   intramuscular), meloxicam (5 mg/kg; subcutaneous) and 0.9% saline
   (1 mL; subcutaneous) pre-operatively. The scalp was removed, and
   bilateral craniotomies were drilled at stereotaxic coordinates ±1 mm
   lateral and −0.45 mm posterior to bregma using standard asepsis
   technique and a standing microscope. A guide cannula was lowered to a
   depth of −2 mm below bregma in each craniotomy and secured using C&B
   metabond (Parkell). A headpost was then secured to the skin using
   Vetbond (3 M) and sealed with C&B metabond. Guide cannulae were plugged
   with homemade pieces of sterile PE 10 tubing (BD Intramedic) secured
   with Kwik-Cast silicone sealant (World Precision Instruments).

   At least two weeks after surgery, cannulae were acutely opened in awake
   mice and aCSF (Tocris, 3525) was infused into each cannula to clear out
   any remaining brain parenchyma, scar tissue and clots. 30 μL of aCSF
   was infused into each cannula over 5 min using a microinjection
   catheter and syringe pump (Kent Scientific Corporation). To prevent
   excessive damage to the ventricular system, infusions of the left and
   right cannula were performed two days apart. Mice without optically
   clear CSF in their guide cannulae following bilateral aCSF infusions
   were not used for photometry experiments. Two days after the final aCSF
   infusion, an optic fiber with metal ferrule (400 μm diameter core;
   multimode; numerical aperture (NA) 0.37; 6.5 mm length; Doric Lenses)
   was placed in each guide cannula and secured with C&B metabond.

   Cannulated wild-type and Ttr^mNeonGreen mice were singly housed in
   standard circadian light–dark boxes and exposed to alternating 12-h
   light and dark periods or constant darkness, depending on the
   experimental paradigm. Fiber photometry of the CSF was performed using
   a FP3001 photometry box (Neurophotometrics) with 415 nm, 470 nm, and
   560 nm LED light sources and a CMOS camera for signal collection (green
   emission channel: 494–531 nm, red emission channel: 586–627 nm).
   Low-autofluorescence fiber optic cables (2 m long; 400 μm diameter
   core; 0.37 NA; Doric Lenses) with three branches were coupled to
   implanted optic fibers with zirconia sleeves (Precision Fiber
   Products). The coupling point was secured with UV-curable optical
   adhesive (Flow-It ALC, Pentron) and shielded with black heat shrink
   tubing. Tethered mice were then allowed to freely move about their
   cages. Freely-moving CSF photometry recording sessions lasted between
   24 and 96 h, during which the CSF was exposed to the following sequence
   of excitation wavelengths: 415 nm (0.2–0.3 mW), 470 nm (0.2–0.3 mW),
   and darkness (background measurement) at 5 Hz. Signal collection was
   controlled using the Bonsai Visual Reactive Program (Open Ephys).

Fiber photometry data analysis

   Photometry data from the green channel at 470 nm excitation were
   preprocessed with a median filter in bins of 10 min. For analyses in
   the first 24 h of each recording, a stretched exponential fit of form
   [MATH:
   <mi>α</mi><mo>+</mo><mi>β</mi><msup><mrow><mi>e</mi></mrow><mrow><mo>−<
   /mo><mi>γ</mi><msup><mrow><mi>t</mi></mrow><mrow><mi>δ</mi></mrow></msu
   p></mrow></msup> :MATH]
   was subtracted from each trace to correct for photobleaching. For
   analyses beyond the first 24 h, no photobleaching correction was
   performed.
   [MATH:
   <mfrac><mrow><mo>△</mo><mi>F</mi></mrow><mrow><msub><mrow><mi>F</mi></m
   row><mrow><mn>0</mn></mrow></msub></mrow></mfrac> :MATH]
   was calculated relative to the first hour of each plot, and F[0] was
   taken as the raw signal prior to photobleaching correction (if it had
   been performed).

Mouse computed tomography (CT) imaging and analysis

   CT scans of the head were obtained to confirm adequate placement of
   guide cannulae in bilateral LVs. Mice in the second post-operative week
   were anesthetized with 1–4% isoflurane (in 100% O[2]) in an induction
   chamber and moved to a heated imaging bed in a SPECT/PET/CT scanner
   (Bruker/Albira). The animal was aligned within the scanner, and CT
   images of the head and neck were acquired. CT images were then imported
   into VivoQuant (Invicro, 2021) for analysis. Because it is difficult to
   distinguish specific brain regions on CT scans, a 3D Brain Atlas Tool
   in VivoQuant was used to manually superimpose a three-dimensional,
   virtual projection of the mouse brain onto our images. Overlap between
   the projected lateral ventricles and bilateral guide cannulae was then
   confirmed.

HRP injection and TEM

   Adult mice were singly housed in cages placed in a circadian cabinet
   controlled lighting environment away from the main mouse colony. The
   lights were programmed to come on at 7 a.m. and turn off at 7 p.m. The
   mice were acclimatized to the boxes for 5 days with ad libitum feeding.
   Mice were weighed and injected intravenously through the tail vein with
   0.5 mg horseradish peroxidase (HRP-type II, Fisher Scientific PI31491)
   per g bodyweight diluted such that each mouse received 100uL per 25 g
   bodyweight. Allow to circulate for 7–10 min, brains were harvested,
   frontal pole was removed, and fixed 1 h in Karnovsky’s fixative, [5%
   glutaraldehyde, 4% PFA, 0.4% CaCl2, in 0.1 M cacodylate buffer - EMS
   (Electron Microscopy Sciences], then fixed in 4% PFA in 0.1 M
   cacodylate at 4 °C overnight while rocking. Brains were coronally
   sectioned in 200 μm vibratome sections. Those sections containing
   ventricle and ChP were selected and processed for DAB (Millipore Sigma
   D5905). Sections were first washed with 20 mM glycine, washed 3 times
   with cold 0.1 M cacodylate, and developed in filter-sterilized 5 mg DAB
   in 9 ml of 0.1 M cacodylate plus ~9 mM H[2]O[2] on ice for 30–45 min
   until dark brown. The portion containing LV ChP was processed,
   sectioned, and imaged at the Conventional Electron Microscopy Facility
   at Harvard Medical School. Tissue was postfixed with 1% osmiumtetroxide
   (OsO4)/1.5% potassium ferrocyanide (KFeCN6) for one hour, washed in
   water three times and incubated in 1% aqueous uranyl acetate for one
   hour. This was followed by two washes in water and subsequent
   dehydration in grades of alcohol (10 min each; 50%, 70%, 90%,
   2 × 10 min 100%). Samples were then incubated in propyleneoxide for one
   hour and infiltrated overnight in a 1:1 mixture of propyleneoxide and
   TAAB Epon (Marivac Canada Inc. St. Laurent, Canada). The following day,
   the samples were embedded in TAAB Epon and polymerized at 60 degrees C
   for 48 h. Ultrathin sections (about 80 nm) were cut on a Reichert
   Ultracut-S microtome, and picked up onto copper grids stained with lead
   citrate. Sections were examined in a JEOL 1200EX Transmission electron
   microscope or a TecnaiG² Spirit BioTWIN. Images were recorded with an
   AMT 2k CCD camera.

Epithelial junction analyses

   Ten images of epithelial junctions were acquired for each animal. N = 3
   animals at each time. Images were acquired at ×2000 for structure.
   Analysis was performed on tiled ×10,000 images reassembled by hand in
   Illustrator (Adobe). A peri-junctional zone was defined as the 2 μm
   wide zone along the lateral edge of 2 ChP epithelial cells.
   Electron-dense HRP-filled vesicles were counted and recorded in FIJI
   ImageJ (NIH). Apical-basal localization was calculated using a custom
   MatLab (Mathworks; R2019b) code as before^[309]124. Tight junction (TJ)
   was defined by the apico-lateral area where plasma membranes make
   complete contact accompanied by continuous, anastomosing network of
   intramembranous particle strands (TJ strands or fibrils). For each TJ,
   10 measurements were made between the 2 plasma membranes within the
   junction that were not in direct contact along the full apico-basal
   extent of the junction. These 10 measurements were averaged to obtain
   the width of each junction. Ten images were analyzed for each animal.
   The average of these 10 averages was plotted for each N. N = 3 animals
   at each time.

Ex vivo ChP 3-day imaging

   After dissection in 1× HBSS, LV ChP from a TTR m:Neon mouse was
   transferred onto 35 mm glass bottom imaging dishes. (MatTek, Cat.
   P35G-1.5-14-C) that had been prepared as follows: briefly, the edges of
   the glass bottom dish were lightly ringed with Silicone (Kwik-sil,
   World Precision Instruments, Item. 600022), and a polycarbonate
   membrane (Whatman, Nucleopore, 13 mm wide, 8.0 mm pore size, Cat.
   110414) with hole cut out of the center was placed on the glass. These
   dishes were kept at room temperature and allowed to cure. Dishes were
   filled with 4 mL of filter-sterilized Lumicycle media exclusive of the
   luciferin (phenol-free DMEM (Sigma D-2902), 10 mM HEPES (pH 8.0), 4 mM
   sodium bicarbonate, 25 mM D-glucose, 1× B27 (Gibco), 4mM L-glutamine,
   Penicillin/Streptomycin). The ChP was flattened onto the glass bottom
   dish and secured to the membrane using 3 M Vetbond. All samples were
   placed in a standard incubator at 37 °C until imaging commenced.
   Explants were imaged in an oxygenated/CO[2]-treated, heated at 37 °C,
   humidified chamber on a Leica DMi8 microscope. Once region was
   identified, tissue was synchronized with 100 nM dexamethasone. Three
   stacks of 40 z-steps were acquired for each timepoint and averaged. One
   set of 3 stacks was acquired every 30 min for 72 h using the HC PL
   fluotarL 20x/0.40 dry objective and Leica-DFC9000GT-[310]VSC07354
   camera. To reduce photobleaching, 460 nm LED power was kept under 18%.

   Prior to analysis, each stack of 3 scans was averaged together, and
   then binned in XY by a factor of 2. A local entropy filter with width
   of 10 pixels was applied to each plane of the resulting stack using the
   MATLAB function entropyfilt, creating an entropy volume with greater
   pixel values in high-detail regions. These high entropy areas
   correspond to regions in which the cell body lattice is in focus. This
   entropy volume was max projected along the Z-axis, and the linear Z
   index is recorded for an in-focus map. This map was smoothed using a
   Gaussian filter with width of 20 pixels, and applied to the original
   volume, resulting in a focused projected image. This image was
   calculated at each 30 min time period for 72 h to make a focused video.

   The focused video was then high-pass filtered and roughly registered,
   first with a rigid registration followed by an affine registration.
   Both registrations were performed using the StackReg algorithm in
   ImageJ. To account for slow-scale warping of the tissue over the course
   of 72 h, a local non-linear registration was then applied by
   calculating the corresponding displacement field between each movie
   frame using the imregdemons algorithm in MATLAB. For each non-linear
   registration, the mean of all frames was used as a reference target.

   This focused and registered image was cropped to the center 50% in both
   X and Y, and the pixel values were rescaled between the 1^st and 99^th
   percentile to normalize the fluorescence change among trials. The
   fluorescent signal was finally calculated by averaging all normalized
   pixel values over time.

Quantification and statistical analysis

   Biological replicates (N) were defined as samples from distinct
   individual animals, analyzed either in the same experiment or within
   multiple experiments, with the exception when individual animal could
   not provide sufficient sample (i.e., CSF), in which case multiple
   animals were pooled into one biological replicate and the details are
   stated in the corresponding figure legends. Statistical analyses were
   performed using Prism (GraphPad) v7 or R (version 3.6.3). Outliers were
   excluded using ROUT method (Q = 1%). Appropriate statistical tests were
   selected based on the distribution of data, homogeneity of variances,
   and sample size. The majority of the analyses were done using One-way
   ANOVA or Student’s two-tailed unpaired t test (when F test failed to
   discover unequal variances) or Welch’s two-tailed unpaired t test (when
   F test discovered unequal variances), except for Fig. [311]3d and
   Supplementary Fig. [312]2c where the analysis was done using
   Kolmogorov–Smirnov tests. F tests or Bartlett’s tests were used to
   assess homogeneity of variances between datasets. Parametric tests (T
   test, ANOVA) were used only if data were normally distributed, and
   variances were approximately equal. Otherwise, nonparametric
   alternatives were chosen. Data are presented as means ± standard
   deviation (SD). If multiple measurements were taken from a single
   individual, data are presented as means ± standard errors of the mean
   (SEMs). Please refer to figure legends for sample size. p values < 0.05
   were considered significant (*p < 0.05, **p < 0.01, ***p < 0.001,
   ****p < 0.0001). Exact p values can be found in the figure legends. P
   values are also marked in the figures where space allows.

Supplementary information

   [313]Supplementary Information^ (7.4MB, pdf)
   [314]Peer Review File^ (197.1KB, pdf)

Acknowledgements