An in vitro model for vitamin A transport across the human blood-brain barrier

Vitamin A, supplied by the diet, is critical for brain health, but little is known about its delivery across the blood-brain barrier (BBB). Brain microvascular endothelial-like cells (BMECs) differentiated from human-derived induced pluripotent stem cells (iPSC) form a tight barrier that recapitulates many of the properties of the human BBB. We paired iPSC-derived BMECs with recombinant vitamin A serum transport proteins, retinol binding protein (RBP) and transthyretin (TTR), to create an in vitro model for the study of vitamin A (retinol) delivery across the human BBB. iPSC-derived BMECs display a strong barrier phenotype, express key vitamin A metabolism markers and can be used for quantitative modeling of retinol accumulation and permeation. Manipulation of retinol, RBP and TTR concentrations, and the use of mutant RBP and TTR, yielded novel insights into the patterns of retinol accumulation in, and permeation across, the BBB. The results described herein provide a platform for deeper exploration of the regulatory mechanisms of retinol trafficking to the human brain.


Introduction 29
Retinoids (Vitamin A and related compounds) are essential micronutrients supplied by 30 the diet that regulate over 500 genes (1) and are involved in vision, embryonic development, cell 31 differentiation, metabolism and brain health (2)(3)(4)(5). Retinoids play critical roles in both brain 32 development and maintenance of brain health. For example, retinoic acid is a key signaling 33 compound that induces neural differentiation and acquisition of unique brain vasculature 34 properties in the prenatal brain (6,7). In the mature brain, retinoids contribute to maintenance of 35 synaptic plasticity and sleep regulation; vitamin A deficiencies can lead to learning and memory 36 deficits and depression of long-term potentiation (8)(9)(10)(11). Vitamin A levels (circulating principally 37 as retinol) naturally diminish with time, and there is mounting evidence that patients with 38 Alzheimer's disease (AD) have lower serum levels of vitamin A than age-matched controls, 39 The blood brain barrier (BBB) is the major site of nutrient exchange between the brain 90 and circulation (52). The primary barrier phenotype is imparted by brain microvascular 91 endothelial cells (BMECs), which are heavily polarized: the luminal (blood-facing) and 92 abluminal (brain-facing) membranes differ significantly in lipid and protein composition (53). 93 The mechanism(s) by which vitamin A is accumulated at the BMEC luminal membrane and then 94 permeated across the BMEC abluminal membrane remains largely unexplored, in large part due 95 to the inherent difficulty in conducting controlled in vivo studies. A diagram of suspected ROH 96 delivery modes across the BBB is shown in Figure 1B. 97 Investigation into the mechanism of ROH transport to the brain would be greatly 98 facilitated by well-characterized in vitro models of the human BBB. Recently, techniques have 99 been developed for reprogramming human-derived cells into an induced pluripotent state 100 (iPSCs) which can be differentiated into cell types that are difficult to obtain from primary 101 Scale bars equal 100 µm. C) Western blots of CRBP1 (red), LRAT (red), and STRA6 (red) 154 confirming antibody specificity. An antibody against TJP1 (green) is used as a BMEC specific 155 loading control in each blot. Polyclonal STRA6 antibody was additionally validated against 156 recombinant GST-tagged STRA6.
recombinant human TTR with a dissociation constant of ~ 250 nM, in close agreement with 164 human serum-derived measurements (58). 165 166 ROH accumulates in iPSC-derived BMEC monolayers from  complexes 168 BMEC monolayers cultured on 96-well plates were exposed to solutions of ROH-RBP or 169 ROH-RBP-TTR prepared at biologically relevant concentrations: 2 µM ROH (typical range in 170 vivo, 1 -2 µM (12,13)), 2 µM RBP (in vivo range, 2 -4 µM (62)), and 4 µM TTR (in vivo range, 171 3 -8 µM (65)). Solutions of ROH-RBP were prepared by overnight equilibration of free lipid 172 ROH (1:20 ratio of 3 H-ROH to unlabeled ROH) and ligand-free (apo) RBP. Solutions of ROH-173 RBP-TTR were prepared by overnight equilibration of ROH-RBP (holo) and TTR. After the 174 desired incubation time for accumulation, cells were washed, lysed, and the tritium signal 175 counted. ROH cellular accumulation, in µmoles of ROH per L cell volume, was calculated from 176 DPM measurements using the manufacturer supplied specific activity of 3 H-ROH, the ratio of 177 3 H-ROH to unlabeled ROH (1:20) and a cell volume of 1.37 x 10 -12 L / cell. Cell volume was 178 estimated by multiplying the average area of a BMEC (402 µm 2 ) by its height (3.4 µm). BMEC 179 area was estimated in two ways: directly from ICC image analysis and by counting the number 180 of cells in a BMEC monolayer after singularization and dividing by the culture dish area. The 181 two methods produced consistent estimates of cell area. BMEC height was estimated by analysis 182 of Z-stack ICC images. We use the term "ROH cellular accumulation" to represent the quantity 183 of all added ROH that becomes cell-associated, recognizing that we did not determine whether 184 any retinol was converted to oxidized metabolites or retinyl esters and we did not differentiate 185 cell-associated radioactivity between internalized versus membrane-associated material after 186 washing. 187 ROH cellular accumulation increased throughout the two-hour experiment at a steady 188 rate with either ROH-RBP or ROH-RBP-TTR ( Figure 3A and Figures S1D, S1F). TTR did not 189 affect ROH cellular accumulation kinetics. After two hours, accumulated ROH cellular 190 concentration was ~100 µM, or about 50-fold higher than the 2 µM ROH medium concentration. 191 This indicates that the BMEC monolayer stores excess ROH and that ROH accumulates against a 192 concentration gradient. High cellular ROH accumulation could be accounted for by a number of 193 established mechanisms, including thermodynamically driven partitioning of the lipophilic ROH 194 into lipid-rich cellular components, binding of ROH by intracellular proteins such as CRPB1,195 and/or storage of internalized ROH as retinyl esters (RE). 196 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  replicates. Measured DPM values were converted to accumulated concentrations using the 209 specific activity of 3 H-ROH, the 3 H- ROH: unlabeled ROH ratio (1:20)  BMECs were cultured on semi-permeable Transwell inserts that allow for sampling of the apical 220 chamber ('blood') and basolateral chamber ('brain') ( Figure 3B). Transendothelial electrical 221 resistance (TEER) was used to confirm the integrity of the BMEC barrier. TEER measures the 222 resistance of the monolayer to an electrical current and correlates with the size of molecules 223 excluded from paracellular transport. For compounds the size of sucrose or retinol (342 Da or 224 accumulation at the one hour time point ( Figure 3C); cellular accumulation of ~60 µM is 239 consistent with the monolayer results at one hour ( Figure 3A) and again shows no statistical 240 difference between ROH-RBP and ROH-RBP-TTR. 241 As shown in Figure 3D, there was a short lag period after which retinoid accumulation in 242 the basolateral chamber increased linearly over the course of the one-hour experiment. 243 Interestingly, although ROH cellular accumulation was not affected by TTR in the apical 244 chamber (Figures 3A and 3C), the basolateral retinoid accumulation was roughly 30% higher 245 when TTR was present in the apical chamber. Basolateral retinoid accumulation after 1 hour was 246 only a small fraction (1-1.5%) of the ROH concentration loaded in the apical chamber. 247 To quantify our data and to confirm basolateral accumulation was due to permeability 248 across the cellular monolayer rather than via paracellular leakage, we used Eq S1 and Eq S2 to 249 calculate the apparent permeability (Peapp) of the BMEC monolayer and Transwell semi-250 permeable insert combined for both sucrose and ROH (Table S3). Sucrose Peapp ranged from 251 0.58 ± 0.03 to 0.69 ± 0.04 x 10 -6 cm/s when mixed with ROH-RBP-TTR or ROH-RBP samples, 252 respectively, in agreement with prior studies using these iPSC-derived BMECs (Pe = 0.57 x 10 -6 253 cm/s) (57) and slightly lower (indicating a tighter barrier) than reported with primary porcine 254 BMECs (Pe = 1 x 10 -6 cm/s, TEER < 1000 Ω x cm 2 ) (67). These results confirm tight barrier 255 integrity in this iPSC-derived model system. Peapp for ROH was 4.8 ± 0.2 x 10 -6 cm/s when 256 supplied by ROH-RBP and 6.6 ± 0.3 x 10 -6 cm/s when supplied by ROH-RBP-TTR. These 257 permeabilities are an order of magnitude higher than those for sucrose and in the same range as 258 that of glucose (Pe = 3.7 x 10 -6 cm/s (57)), a critical nutrient for the brain. This analysis indicates 259 that ROH is indeed transported transcellularly, and that inclusion of TTR increases the ROH 260 permeation rate by ~30%. To our knowledge, this is the first report of permeability 261 measurements for RBP-bound ROH across a BBB-like monolayer, as well as the first reported 262 indicator that TTR may play a role in increasing permeation of ROH across the BBB. 263

264
Free ROH cellular accumulation is bulk fluid concentration-dependent 265 In the presence of RBP and/or TTR, ROH partitions between free and protein-bound 266 states. Although ROH circulating with RBP is thought to be the predominant mode of vitamin A 267 delivery to cells and tissues, there is evidence that free ROH readily partitions into cell 268 membranes (30-33). The estimated concentrations of free and protein-bound ROH at our 269 experimental conditions were calculated from the known equilibrium dissociation constants and 270 total concentrations of ROH, RBP and TTR (Table 1). 271 272 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023

281
For RBP, about 20% of ROH is free (0.4 µM free versus 1.6 µM bound to RBP), while 282 when both RBP and TTR are present, only about 7% (0.14 µM) of ROH is estimated to be free. 283 The relative contribution of free ROH to overall cellular accumulation and permeation across 284 barriers is not known and has typically not been accounted for in other studies of RBP-mediated 285 ROH delivery to cells (41,43,44,49,51,64,70,71), although there is evidence cells respond 286 differently to free ROH as compared to . 287 We used our experimental system to measure the free ROH concentration-dependent 288 cellular accumulation and to compare the data to cellular accumulation data from protein-bound 289 ROH. BMEC monolayers were exposed to ROH at three concentrations: 2 µM (physiological), 290 0.4 µM (to approximate the free ROH concentration in ROH-RBP solutions), and 0.1 µM (to 291 approximate the free ROH concentration in ROH-RBP-TTR solutions). Cell-associated ROH 292 increased over the two-hour time course of the experiment, with total accumulation a strong 293 function of fluid-phase ROH concentration ( Figure 4A and Figure S1A-C). The accumulated 294 cellular ROH concentration after 2 h was nearly two orders of magnitude higher than the fluid-295 phase ROH concentration, consistent with the data observed for protein-bound ROH (Figure 296 then started to increase further at approximately 90 minutes. In contrast, at 2 µM fluid-phase 305 ROH, cellular concentration increased continuously over the two-hour experiment, reaching 306 ~320 µM (~160 µM/µM cell/fluid concentration ratio). 307 308 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023.  We evaluated whether these data could be described by a simple kinetic model, where we 331 assume the bulk concentration remains constant (see Table S5). The strong concentration 332 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 dependence suggested a partitioning model (akin to solvent:solvent partitioning) as a better 333 descriptor than a receptor-ligand binding model. If cf = bulk fluid concentration (µM), ccell = 334 accumulated cellular concentration (µM) at any time t, k1 = first-order rate constant (min -1 ), Kp = 335 partition coefficient (µM cell/ µM fluid), then a simple model is: 336 Eq. 1 337 However, given the apparent intermediate plateau for the 0.4 µM free ROH sample, as 338 well as the observation that the ratio ccell/cf increases with increasing fluid-phase concentration, 339 we hypothesized that a secondary uptake mechanism is triggered after a lag time, tlag, that 340 corresponds to crossing an intracellular ROH threshold, ccell*: We fit the data by Eq. 1 ( Figure S2) or Eq. 2 ( Figure 4B) using least-squares regression. 345 Our analysis indicates that including the secondary uptake mechanism provides a significantly 346 better description of the experimental data ( Figure S2 and Table S5), providing support for the 347 hypothesis of a biphasic response at higher fluid ROH concentrations. A possible explanation is 348 that loading of CRBP1 with ROH to its capacity (estimated from our modeling to be ccell* ~ 36 349 µM) triggers initiation of a secondary storage mechanism, such as retinyl ester synthesis, to 350 handle additional ROH cellular accumulation. Indeed, biphasic ROH uptake from RBP in Sertoli 351 cells has been reported; the accumulated ROH at the plateau in that study corresponded to the 352 intracellular CRBP1 concentration (72). 353 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Free ROH permeates across the BBB-like barrier with kinetics proportional to apical 355 concentration and not cellular concentration 356 Permeation of protein-free ROH at 0.1 µM, 0.4 µM or 2 µM in the apical chamber was 357 measured using the Transwell setup. To ensure barrier integrity, TEER and 14 C-sucrose 358 permeability were monitored as described previously. At the conclusion of each experiment, 359 BMEC monolayer lysate was collected to test for closure of the mass balance; at least 90% of 360 3 H-ROH and at least 96% of 14 C-sucrose radioactivity was recovered (Table S1 and S2). We 361 first confirmed that cellular accumulation in this setup was consistent with that in the monolayer 362 by measuring radioactivity in the lysate at the end of the permeation experiment ( Figure 4C). 363 As shown in Figure 4D, basolateral retinoid accumulation rate increased with increasing apical 364 ROH concentration. We calculated Peapp of ROH to range from 3.8 ± 0.4 to 7.7 ± 0.5 x 10 -6 cm/s 365 as the apical ROH concentration increased from 0.1 to 2 µM. These values indicate that ROH is 366 about 10-fold more permeable than sucrose (0.43 ±0.03 to 0.58±0.03 x 10 -6 cm/s) and are in 367 good agreement with data reported for free ROH (supplied at ~10 -8 M) in primary porcine 368 BMECs (Pe = 4.1 ± 0.7 x 10 -6 cm/s) (67). The permeabilities are similar to those we calculated 369 for ROH-RBP or ROH-RBP-TTR, demonstrating that RBP is not required for ROH transit 370 across the BBB-like barrier. It is important to note that Peapp should be independent of the apical 371 ROH concentration for a single permeation mechanism. The observation that Peapp increases 372 with increasing fluid phase ROH concentration supports the hypothesis that higher fluid phase 373 ROH concentrations trigger secondary transport mechanisms in BMECs or that free ROH is 374 processed intracellularly into different retinoid forms, such as retinyl esters. 375 376 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10. 1101/2023 To understand the concentration dependence of ROH permeation, we normalized the 377 basolateral ROH concentration by the initial apical ROH concentration or by the accumulated 378 cellular ROH concentration (Figures 4E and 4F). The data collapse to a single curve in Figure  379 4E but not in Figure 4F, demonstrating that transport of retinoid across the BMEC monolayer to 380 the basolateral chamber is more tightly coupled to the apical chamber ROH concentration rather 381 than to the accumulated concentration of ROH within the cell barrier. 382

RBP and TTR markedly reduce ROH cellular accumulation compared to protein-free ROH, but 384 permeability across the BBB-like barrier is similar for protein-bound and protein-free ROH 385
We asked whether cellular ROH accumulation kinetics differed between free ROH and 386 protein-bound ROH when supplied at the same initial total ROH concentration. Accumulation 387 kinetics for 2 µM free ROH ( Figure 4A) are compared to accumulation kinetics for 2 µM ROH-388 RBP ( Figure 3A) in Figure 5A. Since the calculated free ROH concentration in the 2 µM ROH-389 RBP sample at these conditions is ~ 0.4 µM ( Table 1), data at 0.4 µM ROH ( Figure 4A) are also 390 included for comparison. ROH cellular accumulation levels from 2 µM ROH-RBP are 391 substantially lower than at 2 µM free ROH, but slightly exceed that of 0.4 µM free ROH alone. 392 For samples containing TTR ( Figure 5B), the free ROH concentration is estimated to be ~0.14 393 µM ( Table 1). ROH cellular accumulation levels from 2 µM ROH-RBP-TTR ( Figure 3A) are 394 markedly lower than those at 2 µM free ROH, but significantly exceed that at 0.1 µM free ROH 395 ( Figure 4A). 396 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; Comparison of basolateral retinoid permeation for free ROH, ROH-RBP and ROH-RBP-TTR 408 samples each containing 2 µM total ROH per sample. 409 410 Cellular uptake of ROH in protein-bound samples could be achieved through two paths 411 operating in parallel: directly from free ROH and/or by delivery from ROH bound to RBP or 412 RBP-TTR. We wondered whether we could estimate the relative importance of these two paths 413 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint by comparing cellular accumulation data from ROH alone versus ROH complexed to RBP or 414 RBP-TTR ( Figure 5C). Specifically, at each time point we normalized the free ROH cellular 415 accumulation by the total ROH cellular accumulation measured for the protein-bound samples 416 (0.4 µM free ROH normalized by ROH-RBP for RBP alone samples and 0.1 µM free ROH 417 normalized by ROH-RBP-TTR for TTR-containing samples -see Table 1). For ROH-RBP, 418 essentially 100% of the initial observed ROH cellular uptake could be accounted for by the 419 contribution of the 0.4 µM free ROH present in the mixture. This percentage drops to 50% after 420 2 hours, as ROH cellular accumulation from ROH-RBP samples grows more quickly than for 0.4 421 µm free ROH alone. This analysis is consistent with the hypothesis that free ROH partitions 422 rapidly into the cell, but then that pathway's contribution diminishes over time as tightly 423 controlled influx / efflux via the ROH-RBP route becomes more dominant. In contrast, ROH 424 cellular accumulation from the 0.1 µM free ROH in ROH-RBP-TTR samples never exceeds 425 ~20% of the total uptake. Thus, although the kinetics of cellular accumulation are virtually 426 identical between ROH-RBP and ROH-RPB-TTR ( Figure 3A), the relative contributions of free 427 versus protein-delivered ROH appear to be quite different ( Figure 5C). This illustrates the 428 importance of careful accounting for all bound and free forms of ROH in mechanistic 429

investigations. 430
Although cellular accumulation was much higher for 2 µM free ROH than when bound 431 to RBP or RBP-TTR, permeation into the basolateral chamber was similar (but not identical) for 432 all three cases, again highlighting that permeation across the BMEC barrier is correlated more 433 strongly with the total apical ('blood') ROH concentration and not with the free apical ROH 434 concentration or the cellular ROH accumulated concentration ( Figure 5D). However, basolateral 435 permeation was not entirely a function of total apical ROH concentration, but also depended on 436 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint whether ROH was presented with RBP alone or in complex with TTR; permeation with RBP 437 alone was lower than for an equivalent total concentration of free ROH, whereas permeation 438 with RBP in complex with TTR was higher than for an equivalent concentration of free ROH. 439 This observation suggests that RBP and TTR might influence ROH cellular trafficking even 440 though the proteins are not themselves internalized, and that the proteins may have subtle 441 differences in the observed effect. 442

RBP and TTR mutants reveal novel insights into mechanisms of ROH cellular accumulation in, 444
and permeation across, BMEC monolayers 445 To further explore the role RBP and TTR play individually in ROH cellular accumulation 446 and permeation, we generated mutants that abrogate wild-type binding interactions. L63R/L64S 447 mutations in RBP (muRBP) alter a loop region at the entrance of the β-barrel of RBP ( Figure  448 6A, adapted from (73)), and this loop region is critical in mediating the binding interaction to 449 TTR. Furthermore muRBP has been previously shown unable to bind to a protein expressed on 450 placental membranes that participates in retinol transport (74); it is suspected, but has not been 451 confirmed, that this protein is STRA6. We produced and purified muRBP, and purity was 452 confirmed using analytical size exclusion chromatography (aSEC) and DEAE anion exchange 453 chromatography ( Figure S3). RBP and muRBP elute in one primary peak on both aSEC and 454 DEAE columns. On aSEC, muRBP elutes earlier than RBP, suggesting the L63R/L64S mutation 455 causes an increase in the apparent volume of the protein. However, refolded ROH-muRBP 456 displays an A330/280 ratio of ~ 1.0 in the presence of excess ROH, indicating normal 1:1 binding 457 stoichiometry (data not shown). Furthermore, ROH binds to muRBP with KD = 110 ± 60 nM, 458 comparable to RBP ( Figure 6B). Interestingly, muRBP fluorescence at saturation is only ~ 50% 459 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint the magnitude of ROH-RBP, which we theorize is due to changes in the β-barrel local 460 environment that alter the efficiency of resonance energy transfer; this hypothesis would be 461 consistent with the apparent larger volume seen by aSEC. To confirm loss of TTR-binding 462 affinity, we utilized a fluorescence anisotropy assay that demonstrates ROH-muRBP does not 463 bind TTR (Figure 6D). 464 465 Figure 6: Characterization of RBP, TTR and mutants. A) PDB Entry 5NU7 (73) was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint

476
To examine the role of TTR binding to RBP on retinol uptake and trafficking, we 477 produced an I84A mutant TTR (muTTR) with reduced affinity for RBP (28,76) ( Figure 6C, 478 adapted from (75)). Fluorescence anisotropy confirmed that recombinant muTTR did not bind 479 measurably to ROH-RBP ( Figure 6D). 480 First, we measured ROH cellular accumulation in BMEC monolayers using muRBP or 481 muTTR and compared the results against RBP or TTR. We found negligible differences in ROH 482 cellular accumulation between ROH-RBP and ROH-RBP-muTTR ( Figure  Surprisingly, accumulation from ROH-muRBP is markedly higher than that of ROH-495 RBP; indeed, accumulation approaches that of free ROH at 2 µM. This was an unexpected result 496 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; given that muRBP binds ROH with similar affinity as RBP, but may be consistent with the 497 hypothesis that muRBP cannot participate in ROH efflux from intracellular ROH stores, and 498 thereby drives a higher net ROH accumulation. 499 We next measured ROH permeability across the BMEC monolayer in the Transwell 500 configuration using muRBP or muTTR with RBP. We confirmed consistency of the monolayer 501 plate and Transwell experiments, showing again that cellular ROH accumulation was 502 significantly higher with muRBP than with RBP ( Figure 7B). Although cellular accumulation 503 was much higher for ROH bound to muRBP than compared to RBP, basolateral retinoid 504 permeation was virtually identical ( Figure 7C). This result is consistent with our other 505 observations that basolateral permeation is primarily correlated with apical and not cellular ROH 506 concentration. 507 Because the I84A TTR mutant does not bind to RBP, we anticipated that ROH-RBP-508 muTTR would demonstrate permeability similar to ROH-RBP alone. However, basolateral 509 permeability was higher in samples containing TTR or muTTR when compared to ROH-RBP 510 alone ( Figure 7C). This surprising result indicates that it is the presence of TTR, and not 511 specifically the binding of TTR to RBP, that is responsible for the higher basolateral permeation 512 of ROH. 513

TTR and muTTR upregulate expression of LRAT 515
It is known that influx of ROH delivered by ROH-RBP through STRA6 triggers an 516 intracellular signaling cascade (46,48). STRA6-mediated transport of ROH is coupled to 517 intracellular CRBP1 (47,49) concentrations, while LRAT plays an important role in storing 518 excess intracellular ROH by enzymatically converting it to retinyl esters. We used our in vitro 519 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint system to explore whether ROH uptake may trigger transcriptional changes in these vitamin A-520 related proteins when treated with ROH and its binding partners. Specifically, we used RT-qPCR 521 to measure changes in expression of STRA6, CRBP1 and LRAT after treatment of BMECs with 522 the ROH preparations listed in Table 1. Primer sequences used are detailed in Table M3. Expression values are normalized to the housekeeping gene, ACTB, and quantified relative to 526 BMECs treated with HBSS alone. ΔΔCq data are presented in box-and-whisker format after log2 527 transformation with the values for each biological replicate displayed individually (N = 4). 528 Statistical analyses were performed in Prism on log2 transformed ΔΔCq values via one-way 529 ANOVA followed by Dunnett's test using a confidence interval of 95%. A) STRA6; B) CRBP1; 530 C) LRAT. 531 532 Neither STRA6 nor CRBP1 expression was affected to a statistically significant degree 533 by addition of ROH, RBP or TTR (Figure 8A and Figure 8B). Nor was there a significant 534 change in LRAT expression when BMECs were exposed to free ROH, ROH-RBP, or ROH-535 muRBP ( Figure 8C). Samples containing either ROH-RBP-TTR or ROH-RBP-muTTR, 536 however, show statistically significant upregulation of LRAT ( Figure 8C). Since TTR binds to 537 RBP but muTTR does not, LRAT upregulation is independent of TTR binding to RBP and is 538 decoupled from the delivery of ROH through STRA6 via RBP. Since LRAT plays a role in 539 managing intracellular ROH inventory, this result could explain why the presence of TTR, and 540 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint not its binding to RBP, increases ROH permeability across the BMEC monolayer ( Figure 7C). 541 Although more detailed investigation is needed, our data are indicative of a heretofore unknown 542 function for TTR in regulating ROH delivery across barriers, independent of its well-known role 543 as a carrier for RBP. 544 545 Discussion 546 Despite the importance of retinoids to brain health, very little is known about the 547 mechanisms by which retinoids enter the specialized cells forming the blood-brain barrier, or 548 how retinoids transit the BBB. Using human iPSC-derived BMECs, as well as recombinant 549 human RBP and TTR, we constructed an in vitro platform that provides a practical means for 550 controlled experimentation and quantitative interrogation of retinol uptake by, and delivery 551 across, the BBB. iPSC-derived BMECs express markers indicative of brain endothelial 552 specificity, including PECAM1, Claudin-5, Occludin, ZO-1 and Glut1. BCRP and MRP1 553 expression provide evidence of efflux transporter activity, a major function of the in vivo BBB. 554 In this work we show that BMECs express STRA6 as well as two other vitamin-A relevant 555 proteins: LRAT and CRBP1, both of which are known to couple with STRA6 for ROH uptake. 556 Recombinant RBP and TTR were produced that are suitable replacements for human plasma-557 derived materials while providing a means to probe the roles of these proteins individually in 558 ROH trafficking and transport via specific mutations. Using this in vitro system, we measure for 559 the first time the rate of ROH transport across a human BBB-like barrier and we show that ROH 560 permeability is similar to that of an essential brain nutrient, glucose. 561 Our results clearly show that cells take up free ROH in a concentration-dependent 562 manner, and that ROH binding proteins are not required for cellular accumulation (Figure 4A). 563 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 After two hours, cell-associated ROH concentrations exceed that in the fluid phase by two orders 564 of magnitude. Since transport of free ROH into the cell is maintained against a strong 565 concentration gradient, this result indicates that iPSC-derived BMECS store ROH primarily in a 566 bound or associated form. This is similar to another blood-barrier, the retinal pigment epithelium 567 (RPE), where 92% of total intracellular ROH is stored as retinyl esters (RE) and 8% as ROH-568 CRBP1 complex (77). 569 The kinetic pattern of BMEC ROH uptake varied with ROH concentration (Figure 4B). 570 The biphasic behavior of ROH uptake in Sertoli cells is similar to what we observed in iPSC-571 BMECs at intermediate concentrations of ROH; the plateau in that study occurred at an 572 intracellular ROH concentration corresponding to the CRBP1 concentration (72). We propose 573 that, at 0.1 µM fluid phase ROH, BMECs maintain an adequate sink for intracellular ROH, 574 possibly through binding to CRBP1. At higher ROH concentrations, ROH accumulates 575 intracellularly beyond the CRBP1 binding capacity, and a secondary storage system is recruited 576 into action. A simple kinetic model was developed that is consistent with this proposed 577 mechanism and provides an estimate of the 'triggering' cellular ROH concentration as ~36 µM. 578 The secondary storage system may proceed through LRAT, which normally esterifies ROH 579 bound to CRBP1 and is regulated by the ratio of bound to unbound CRBP1; however, in periods 580 of excess free ROH, LRAT can act on free ROH directly (78,79). Further study is needed to 581 validate this hypothesis. 582 Not only can free vitamin A be taken up by BMECs, but it also can be transported across 583 the barrier without requiring RBP or TTR (Figure 4D), see Figure 1B. Notably, basolateral 584 ('brain') accumulation was directly proportional to the apical ('blood') ROH concentration, and 585 not to the cellular concentration. This suggests that cells at the BBB store excess ROH and 586 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint carefully regulate retinoid delivery to the 'brain' in response to blood concentrations, and/or that 587 protein-delivered ROH has greater ability to access trans-cellular trafficking pathways than ROH 588 taken up as free lipid. 589 RBP carries out two key functions: it sequesters ROH in the blood in order to maintain 590 free ROH concentrations at sub-toxic levels, and it delivers ROH to cells via STRA6. We tested 591 whether ROH accumulation in, or retinoid transport across, BMECs was enhanced by RBP. 592 Notably, complexation of ROH with RBP slows cellular accumulation considerably relative to 593 what occurs in the absence of RBP at an equivalent ROH concentration (compare ROH-RBP 594 versus 2 µM free ROH, Figure 5A). Interpretation of this result requires consideration of the fact 595 that there is ~0.4 µM free ROH in the ROH-RBP preparation, and a better comparison therefore 596 should account for the contribution of the 0.4 µM free ROH present in the ROH-RBP 597 preparation. We hypothesize that cellular uptake of ROH occurs by two parallel paths, from free 598 ROH and from ROH-RBP. A direct comparison of the kinetics (Figure 5C) leads to our 599 hypothesis that free ROH rapidly partitions into the cell, whereas the RBP-mediated cellular 600 accumulation is slower; after two hours, however, about half of the total ROH cellular 601 concentration may be accounted for by each of the pathways. 602 Experiments with the L63R/L64S mutant RBP provide further insight into the role of 603 RBP in regulating ROH trafficking. Cellular uptake of ROH is significantly faster with muRBP 604 compared to RBP, and accumulation reached levels approaching that of 2 µM free ROH. This 605 was an unexpected result, since muRBP binds ROH with equal affinity to RBP ( Figure 6B). As 606 illustrated in Figure 1B, ROH can cross cellular membranes through multiple mechanisms: by 607 passage of free ROH through the lipid bilayer, by passage through cell-surface proteins, or by 608 STRA6-mediated release of ROH from ROH-RBP. Importantly, RBP-mediated transport 609 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint through STRA6 is bidirectional, with both influx and efflux depending on availability of 610 intracellular apo-CRBP1 and extracellular apo-RBP, respectively (43). Furthermore, ROH efflux 611 requires binding of apo-RBP to STRA6 (43,49), a mechanism available to RBP but presumably 612 not muRBP (74). We postulate that cellular accumulation is therefore the net sum of three steps: 613 uptake of free fluid-phase ROH, uptake of ROH via delivery from holo-RBP, and efflux of 614 cellular ROH via uptake by extracellular apo-RBP. The substantially higher cellular 615 accumulation for muRBP compared to RBP provides support for the hypothesis that efflux of 616 ROH is a critical regulatory component for BMEC intracellular ROH levels and that this 617 pathway is inoperable with muRBP. Additional support for this hypothesis would require direct 618 characterization of the binding of RBP and muRBP to our iPSC-derived BMECs, a subject for 619 future research. Regardless, these data suggest that the most important role of RBP during times 620 of retinol abundance is to mediate efflux of excess intracellular ROH, thereby maintaining 621 intracellular ROH at sufficient, but not excessive, concentrations. 622 Despite the large differences in ROH cellular accumulation with 2 µM free ROH versus 623 the same quantity incubated with RBP or muRBP, the ROH permeation kinetics were very 624 similar ( Figure 7C). This result suggests that BMEC monolayers regulate transport of ROH 625 across the barrier by 'sensing' the apical ('blood') ROH concentration. A plausible mechanism 626 for achieving 'sensing' is via the intracellular ratio of ROH-bound CRBP1 and ligand-free apo-627 CRBP1. Apo-CRBP1 promotes retinyl ester (RE) hydrolysis (79,80) and influx of ROH through 628 STRA6 (43), while ROH-CRBP1 promotes RE synthesis via LRAT (78) and efflux of ROH 629 through STRA6 (43). Efflux of ROH through STRA6 requires availability of ligand-free RBP 630 (apo-RBP) in the blood. Therefore, the ratio of intracellular ROH-CRBP1 to apo-CRBP1 is 631 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint directly coupled via STRA6 to blood ROH through the ratio of extracellular ROH-RBP to apo-632

RBP. 633
TTR's primary role in ROH transport is generally believed to be in binding to RBP and 634 preventing loss of the small protein through the kidney (28). ROH is more buried in the ROH-635 RBP-TTR complex compared to ROH-RBP (28), and cross-linking studies show evidence of an 636 ROH-RBP-STRA6 complex but not of an ROH-RBP-TTR-STRA6 complex (51); taken 637 together, these observations suggested that dissociation of ROH-RBP from TTR is required for 638 RBP-mediated transfer of ROH to STRA6. Indeed, we observed that ROH cellular accumulation 639 with ROH-RBP-TTR was indistinguishable from ROH-RBP (Figure 3A), which at first glance 640 would indicate that TTR does not play a direct role in ROH uptake into the BMEC monolayer 641 (see Figure 1B, iii). However, there are significant differences in the ROH distribution between 642 ROH-RBP and ROH-RBP-TTR delivery modalities that may suggest a more complicated 643 scenario: first, the free ROH concentration is ~ 4x lower when both RBP and TTR are present 644 (~0.14 µM, see Table 1) than compared to RBP alone (0.4 µM, see Table 1); second, the 645 concentration of ROH-RBP differs by an order of magnitude between RBP and TTR containing 646 samples (0.18 µM, see Table 1) and the RBP alone samples (1.6 µM, see Table 1). The 647 comparative analysis shown in Figure 5C indicates that direct uptake of free ROH in ROH-648 RBP-TTR mixtures contributes much less to the overall cellular accumulation than when 649 compared to ROH-RBP mixtures alone. Since there is no evidence of direct binding of ROH-650 RBP-TTR to STRA6, we postulate that ROH-RBP in ROH-RBP-TTR mixtures must play a 651 much larger role in ROH cellular influx than when ROH-RBP is presented alone. This may have 652 significant physiologic ramifications; STRA6-mediated JAK/STAT signaling is activated 653 specifically by ROH influx via ROH-RBP (see Figure 1B, ii) (47); furthermore, this signaling 654 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint cascade requires ROH to be delivered directly from RBP, as neither RBP nor ROH alone induce 655 the cascade (46). 656 When TTR was added to the apical chamber, we observed higher permeation into the 657 basolateral chamber, and a higher Peapp than that of free ROH or ROH-RBP at the same total 658 apical ROH concentration ( Figure 7C). Surprisingly, this effect was seen with both TTR and 659 muTTR, despite the fact that muTTR does not bind to ROH-RBP. This result indicates that it is 660 the presence of TTR, and not the binding of TTR to RBP, that is responsible for the higher 661 permeation of ROH. To gain further insight, we looked for changes in expression of STRA6, 662 CRBP1 or LRAT. No statistically significant changes in STRA6 or CRBP1 expression were 663 detected after two hours of incubation. With LRAT, however, we saw a statistically significant 664 increase in expression in BMECs exposed to either TTR or muTTR, but not to ROH or ROH-665 RBP alone. This is a novel finding that suggests TTR plays an important role in regulating 666 retinol trafficking and transport across barriers. Moreover, this regulatory activity does not 667 require TTR to be complexed to RBP. If TTR signaling directly increases expression of LRAT, 668 concomitant enhanced production of RE could serve as the substrate for basolateral efflux, the 669 mechanistic basis of which is wholly unknown. Further studies are required to tease out this 670 unexpected role of TTR in retinol processing. 671 Taken together, our results demonstrate the utility of our in vitro BBB model constructed 672 from iPSC-derived BMECs and recombinant wild-type and mutant RBP and TTR for studies of 673 retinol trafficking across the BBB. Several novel findings include: (1) accumulation of ROH in 674 the cells of the BBB is a strong function of the delivery mode (free or protein-bound), while 675 permeation across the BBB is mostly independent of delivery mode, (2) retinol permeation rates 676 across the BBB are similar to that of glucose, another essential brain nutrient, (3) efflux of ROH 677 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 through STRA6 to apo-RBP in the serum may be an underappreciated route for controlling 678 intracellular accumulation in times of retinol abundance, (4) TTR upregulates LRAT expression 679 and influences ROH transport to the brain using mechanisms that are independent of its RBP-680 binding role. We highlight the importance of using wild-type and mutant RBP and TTR, as well 681 as careful accounting for the distribution of ROH between free and protein-bound states, in any 682 mechanistic investigation of retinol trafficking, permeability or STRA6-mediated signaling. 683 By leveraging the renewable, scalable and genetically-manipulable features of human 684 iPSC-derived BMECs, the roles of key cellular proteins involved in retinoid accumulation, 685 metabolism and permeation across the BBB could be directly explored. Recent advances in 686 CRISPR-mediated perturbation methods, such as knockout (CRISPRko), interference (CRISPRi) 687 or activation (CRISPRa), can be used to map the roles of STRA6, LRAT and/or CRBP1 in 688 retinoid processing at the BBB. Results reported here lay the groundwork for more detailed 689 investigations into mechanisms of retinol transport into the brain, which are expected to yield 690 greater insights into retinol's role in supporting brain health and to provide novel approaches for 691 treatment of retinol dysfunction in neurodegenerative disease. 692 693

Induced pluripotent stem cell (iPSC) differentiation to brain microvascular endothelial-like cells 695
(BMECs) 696 IMR90-4 iPSCs (WiCell, Madison, WI) were cultured on Matrigel-coated 6-well plates 697 and supplemented daily with E8 medium (Stem Cell Technologies) as described (81). iPSCs 698 were passaged in clumps at ~ 70% confluency every 3 -5 days by dissociation with Versene 699 (Life Technologies) at typical ratios between 1:6 and 1:12. To initiate differentiation into 700 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; BMECs, iPSCs at ~ 70% confluency were dissociated and singularized by treatment with 701 Accutase (Life Technologies) for 7 -10 minutes and then diluted into fresh E8 media (1:4 v/v). 702 Cells were counted on a hemocytometer, and then centrifuged at 200 x g for 5 minutes. The 703 supernatant was aspirated and the pellet re-suspended in fresh E8 medium supplemented with 10 704 µM ROCK inhibitor (Tocris R&D Systems). Re-suspended cells were seeded at a density 705 between 7,500 -12,500 cells / cm 2 on fresh Matrigel-coated plates (Day -3). ROCK-706 supplemented E8 media was aspirated and replaced with fresh E8 media (without ROCK 707 inhibitor) 24 hours later to promote iPSC colony formation. Cells were subsequently expanded 708 for 48 hours with daily E8 media replacement until reaching the optimal density of 30,000 cells / 709 cm 2 as described previously (82) was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 concentration of collagen and fibronectin, respectively, was 400 µg/mL and 100 µg/mL. Each 724 Transwell filter was coated with 200 µL of 4:1:5 solution. For all other plasticware, 4:1:5 725 solution was further diluted 5-fold in sterile water. On Day 8, cells were singularized with 726 Accutase for 30 -45 minutes, re-suspended in EC +/+ medium and plated onto the 4:1:5 727 collagen IV / fibronectin-coated plasticware prepared fresh the day before at a density of 1 × 10 6 728 cells / cm 2 for 1.12 cm 2 Transwell-Clear permeable inserts, at a density of 250,000 cells / cm 2 on 729 6/12-well tissue culture polystyrene plates, or at a density of 1 × 10 6 cells / cm 2 for 24/48/96-well 730 tissue culture polystyrene plates. On Day 9, EC +/+ medium was replaced with EC +/-medium 731 without retinoic acid to promote barrier tightening. At least one Transwell was seeded per 732 differentiation to monitor transendothelial electrical resistance (TEER) as a measure of BMEC 733 quality. TEER was measured every 24 hours after subculture on Day 8 to confirm barrier 734 tightness. Resistance was recorded using an EVOM ohmmeter with STX2 electrodes ( and fixed with either 100 % Methanol or 4 % paraformaldehyde for 10 -15 minutes at room 744 temperature. After fixation, the fixing agent was aspirated and the cells were washed three times 745 in immediate succession with DPBS. After washing, cells were incubated for 60 minutes at room 746 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint temperature in blocking buffer (10 % v/v goat serum in DPBS) before overnight incubation at 4 747 °C with primary antibodies diluted in blocking buffer (Table M1). 748 After primary incubation, cells were washed three times at 5 minutes each with DPBS 751 before incubation for 1 hour at room temperature in the dark with secondary antibodies in 752 blocking buffer (goat anti-rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody conjugated to 753 Alexa Fluor 488 (ThermoFisher, 1:200 dilution, Cat # A-11008) or goat anti-mouse IgG1 Cross-754 Adsorbed Secondary Antibody conjugated to Alexa Fluor 488 (ThermoFisher, 1:200 dilution, 755 Cat # A-21121)). After secondary antibody incubation, cells were immediately stained for 15 756 minutes at room temperature in the dark with Hoechst nuclear count stain (ThermoFisher) 757 diluted 1:5000 in DPBS. Cells were washed three times and visualized in fresh DPBS on a Nikon 758 Ti2 epifluorescence microscope with a 20 X objective. Images were analyzed with FIJI software. 759

Western blot analysis for LRAT, CRBP1 and STRA6 expression 760
BMECs were singularized on Day 8 and seeded on 6-well plates coated with 4:1:5 761 solution. On Day 10, BMECs were rinsed once with DPBS (ThermoFisher) and lysed for 15 762 minutes at 4 °C using ice-cold radioimmunoprecipitation assay buffer (Rockland) supplemented 763 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 with protease inhibitor cocktail (Pierce) according to the manufacturer's recommendation. 764 Lysate, including cell membrane debris, was collected via scraping and centrifuged at max speed 765 for 10 minutes at 4 °C. Supernatant was collected and the protein concentration quantified by 766 bicinchoinic acid assay according to the manufacturer protocols (Pierce). Lysate was prepared 767 before overnight incubation at 4 °C with primary antibodies diluted in blocking buffer (Table  778 M2). 779 After primary incubation, membranes were washed three times at 5 minutes each in 782 TBST before incubation for 1 hour at room temperature in the dark with goat anti-rabbit IgG 783 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; IRDye 680RD (Li-Cor, 1:5000 dilution, Cat # 926-68071) and donkey anti-mouse IgG IRDye 784 800CW (Li-Cor, 1:5000 dilution, Cat # 926-32212) secondary antibodies in blocking buffer. 785 Membranes were washed three times for 5 minutes each in TBST and imaged using a Li-Cor 786 Odyssey Imager. 787

Preparation of retinol, RBP and TTR 788
All-trans retinol (Sigma) and alpha-tocopherol (Sigma), which served as an antioxidant 789 stabilizer, were dissolved in ethanol in equimolar concentrations and stored at -80 °C. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 phosphate buffered saline (PBS) via centrifugal filtration. Concentrations were quantified via 807 absorption using an extinction coefficient of 40,400 M -1 cm -1 at 280 nm. ROH-RBP and ROH-808 muRBP used for experiments were prepared as equimolar solutions 24 -48 hours prior to use. 809 Critically, confirmation of ROH binding was determined by absorption at 330 nm. The 810 equilibrium dissociation constants for ROH to RBP and ROH to muRBP were both measured 811 using fluorescence spectroscopy as described (58) The protein solution was applied to a chitin affinity column and allowed to refold on column. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 on Day 10 for retinol accumulation assays, performed at 37 °C on a shaker. BMECs from the 830 same differentiation were seeded in parallel on Transwells to confirm quality via TEER 831 measurements. Briefly, on Day 10 of the differentiation, cell media was aspirated from all wells 832 and replaced with solutions composed of HBSS (ThermoFisher) and ROH, ROH-RBP, ROH-833 muRBP, ROH-RBP-TTR or ROH-RBP-muTTR. Unlabeled ROH was mixed one hour prior to 834 use with a 3 H tracer prepared identically at a target final concentration of 5% 3 H-ROH. The 835 actual concentration was quantified by liquid scintillation and tracer percentage adjusted 836 accordingly. For each experiment, one 96-well plate was subdivided into 25 triplicate groups and 837 loaded with the various ROH preparations. Wells were aspirated serially in triplicate every 5 838 minutes over the course of 2 hours, and each well was considered a biologically distinct 839 replicate. Immediately after aspiration, each well was washed twice with HBSS and allowed to 840 dry. All wells were lysed simultaneously by addition of 100 µL of ice-cold 841 radioimmunoprecipitation assay (RIPA) buffer per well and incubated at 4 °C for 10 minutes. 842 Following lysis, each 100 µL lysate / RIPA mixture was placed into an individual liquid 843 scintillation counting vial and measured for 3 H DPM. Briefly, vials were diluted with 10 mL of 844 UltimaGold (PerkinElmer), shaken vigorously, and counted immediately three times for 5 845 minutes each on a Tri-Carb 2900TR Liquid Scintillation Counter (PerkinElmer). The average 846 DPM from the three technical replicate readings was utilized as the readout. Tritiated samples 847 were counted using a preset region of 0 -18.6 keV, while carbon-14 samples, where required, 848 were counted simultaneously using a preset region of 0 -156 keV. Self-normalization and 849 calibration (SNC) was performed using external standards prior to each data run. DPM were 850 converted to cellular concentrations by using the specific activity of the tritiated retinol and an 851 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  QIAshredder columns (Qiagen). Samples loaded on spin columns were treated with DNase 865 (Qiagen) to digest genomic DNA prior to RNA purification. RNA was eluted in molecular 866 biology grade water (Corning) and quantified by UV-vis absorbance using an Eppendorf 867 BioSpectrometer. Sample quality was confirmed by a clean absorbance scan with an A260/280 ratio 868 of > 1.8. 200 ng of purified RNA was immediately reverse transcribed at 37 °C for one hour in 869 an S1000 Thermal Cycler (BioRad) via the OmniScript RT Kit (Qiagen) according to the 870 manufacturer's instructions using RNase Out (ThermoFisher) and Oligo(dT)12-18 primers 871 (ThermoFisher). cDNA was used immediately or stored at 4 °C for no more than 24 hours prior 872 to qPCR. 873 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint qPCR samples were prepared using 2X PowerUP SYBR Green (ThermoFisher), 10 ng of 874 cDNA template prepared previously and PCR primers diluted in molecular biology grade water 875 as described in Table M3. 876 Samples were loaded onto qPCR skirted plates (Agilent) and placed into an AriaMx 879 Real-time PCR System (Agilent). Each run was initiated with a hot start at 95 °C for 15 minutes 880 followed by 50 cycles consisting of a 15s denaturation step at 95 °C followed by a 60s 881 amplification step at 60 °C. Fluorescence was analyzed after each cycle. After 50 cycles, the 882 PCR product was melted and melt curves were inspected after each run to ensure only one peak 883 was observed, corresponding to the appropriate PCR product as determined by agarose gel 884 electrophoresis. Each condition analyzed contained 4 biological replicates with three technical 885 replicates for each gene. An arithmetic mean Cq value was calculated from the technical 886 replicates for each gene, and this value for BMECs treated with HBSS alone was used as the 887 reference to calculate ΔCq values for each biological replicate of each gene for each ROH 888 delivery condition. ΔCq values were converted to normalized expression quantities using an 889 assumed 100% efficiency and ACTB as a housekeeping gene. Statistical analyses were 890 performed on log2-transformed relative fold change quantities using one-way ANOVA followed 891 by Dunnett's test with an assumed confidence interval of 95%. Log2-transformed relative fold 892 change data are presented graphically in box-and-whisker format. 893

894
Retinol BBB permeability assays 895 BMEC monolayers cultured on collagen / fibronectin 12-well Transwells were prepared 896 as described. On Day 10 of the differentiation, TEER was optimal and BMECs were used for 897 permeability assays. 550 µL of ROH-containing solution was applied in triplicate to the apical 898 chambers and 1500 µL of HBSS was added into each basolateral chamber. ROH preparations 899 were spiked with 14 C-sucrose (PerkinElmer) as a paracellular transport control. Samples 900 collected from the apical and basolateral chambers over the course of the experiment totaled 10% 901 of the bulk volume, and liquid removed from basolateral chambers was immediately replenished 902 with the same volume of fresh HBSS. At t = 0, samples were collected from both apical and 903 basolateral chambers. Plates were then placed at 37 °C on a shaker. Samples were collected from 904 the basolateral chambers every 15 minutes. At t = 60 minutes, an additional sample was also 905 collected from the apical chamber. Once removed, each sample was placed into a scintillation 906 vial for counting. TEER was measured after the final sample collection at 60 minutes to verify 907 integrity of the monolayer. At the conclusion of the assay, solutions in both the apical and 908 basolateral compartments were aspirated and 100 µL of ice-cold RIPA buffer was added directly 909 to the BMEC monolayer to lyse the cells. After incubation at 4 °C for 10 minutes, cell lysate was 910 scraped and collected for scintillation counting. Concentrations of tritium in the apical and 911 basolateral chambers were calculated using the specific activity and well volume. Concentrations 912 of tritium associated with the BMEC monolayers were calculated using the specific activity and 913 the estimated volume of a single BMEC cell. 914 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023

Eq (S2)
2 566 is the mean apparent permeability (N = 4 replicates) of the BMEC monolayer and Transwell filter combined calculated by Eq (S2), where 9 is the Transwell filter area and ; < is the fitted linear slope of the clearance volume ( & ) as a function of time. For samples displaying a lag-phase, slope was calculated only from the linear segment. & values were calculated by Eq (S1), where ) is the volume of the donor chamber and +,-and ),-/ 01 234 are the CPM signal in the acceptor chamber at time and the CPM signal in the donor chamber at time = 60 , respectively.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint 1 in gray). If this model were correct, the data at all free ROH concentrations would collapse onto a single curve described by Eq. 1.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ; https://doi.org/10.1101/2023.04.11.536348 doi: bioRxiv preprint pre-equilibrated with AEX Buffer A. The sample was allowed to adsorb for 10 minutes, then reequilibrated for 10 minutes with AEX Buffer A at a flow rate of 1.0 mL/min. A step salt gradient (dotted line) was applied by mixing AEX Buffer A with high salt AEX Buffer B (25 mM Tris, 1 M NaCl, 1 mM EDTA, pH 8.0) at 1.0 mL/min. . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 12, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023