The solute carriers ZIP8 and ZIP14 regulate manganese accumulation in brain microvascular endothelial cells and control brain manganese levels

Manganese supports numerous neuronal functions but in excess is neurotoxic. Consequently, regulation of manganese flux at the blood–brain barrier (BBB) is critical to brain homeostasis. However, the molecular pathways supporting the transcellular trafficking of divalent manganese ions within the microvascular capillary endothelial cells (BMVECs) that constitute the BBB have not been examined. In this study, we have determined that ZIP8 and ZIP14 (Zrt- and Irt-like proteins 8 and 14) support Mn2+ uptake by BMVECs and that neither DMT1 nor an endocytosis-dependent pathway play any significant role in Mn2+ uptake. Specifically, siRNA-mediated knockdown of ZIP8 and ZIP14 coincided with a decrease in manganese uptake, and kinetic analyses revealed that manganese uptake depends on pH and bicarbonate and is up-regulated by lipopolysaccharide, all biochemical markers of ZIP8 or ZIP14 activity. Mn2+ uptake also was associated with cell-surface membrane presentation of ZIP8 and ZIP14, as indicated by membrane protein biotinylation. Importantly, surface ZIP8 and ZIP14 biotinylation and Mn2+-uptake experiments together revealed that these transporters support manganese uptake at both the apical, blood and basal, brain sides of BMVECs. This indicated that in the BMVECs of the BBB, these two transporters support a bidirectional Mn2+ flux. We conclude that BMVECs play a critical role in controlling manganese homeostasis in the brain.

Manganese is an essential nutrient utilized as a cofactor for various enzymes, e.g. glutamine synthetase, superoxide dismutase, and glycosyltransferases (1). However, imbalance in manganese homeostasis caused by environmental overexposure or dysregulation of manganese trafficking has been correlated with a variety of neuropathologies. Overexposure can lead to manganese deposits in the basal ganglia, the region responsible for voluntary muscle coordination resulting in manganism, hypermanganesemia, and Parkinson's disease (2). Manganese overload at the cellular level promotes lipid, protein, and DNA oxidation and induces mitochondrial stress (3). Clinical investigations have identified patient mutations in putative manganese transport proteins, providing clues to some of the proteins involved in maintaining normal manganese levels (4 -8). However, how these proteins function in the management of manganese within specific organs and cell types has not been fully elucidated. The goal of the research described herein was to identify the transport proteins involved in Mn 2ϩ uptake in the brain microvascular endothelial cells that form the barrier between the circulation and the brain's interstitium.
The molecular mechanisms by which manganese enters into the brain from the systemic circulation have not been wellcharacterized. Crossgrove et al. (9,10) examined the general features of manganese uptake via perfusion studies in isolated brains. These authors demonstrated that the rate of manganese transfer from the blood into the brain was faster than could be described by simple diffusion; they demonstrated also that Mn 2ϩ , Mn 2ϩ -citrate, and Mn 3ϩ -transferrin were all substrates for the accumulation of brain manganese. Fitsanakis et al. (11) observed that manganese accumulation in rat BMVECs 2 was influenced by temperature and pH, was energy-dependent, and was inhibited by other divalent metal ions. In this work we identify those proteins responsible for Mn 2ϩ uptake into the brain, demonstrate their membrane presentation, and examine factors that modulate their activity in Mn 2ϩ uptake.
The blood-brain barrier is a highly selective barrier, primarily composed of brain microvascular endothelial cells surrounded by a basement membrane and pericytes, which regulate microvessel blood flow. BMVECs are also in direct contact with astrocyte end-feet and neurons within the brain parenchyma. Unlike most blood vessels in the body, brain capillaries are deprived of fenestrations; rather, BMVEC form tight junctions much like those formed by epithelial cells facing the intestinal lumen (12). Thus, nutrients entering the brain from the blood must do so by transcellular-mediated uptake and efflux (13). Significantly, little attention has been paid to the alterna-tive trajectory, from the basal, abluminal space into the apical, blood side of this barrier.
Under normal conditions, manganese levels in the blood range from 4 to 15 g/liter or 75 to 300 nM (14). Divalent manganese is bound to small ligands such as HCO 3 Ϫ and citrate and loosely bound to larger proteins like albumin and macroglobulin (15). Trivalent manganese complexes with pyrophosphate or transferrin as chelating agents are found, also; the latter supports receptor-mediated uptake of Mn 3ϩ (16,17). A number of membrane proteins have been implicated in cellular Mn 2ϩ uptake in mammalian cells, including the citrate shuttle, plasma membrane calcium channels, DMT1 (divalent metal transporter 1), ZIP8 (solute carrier family 39 member 8, SLC39A8), and ZIP14 (solute carrier family 39 member 14, SLC39A14); the latter three are known to be promiscuous divalent metal importers (18). As noted, manganic transferrin is transported into/out of mammalian cells via the transferrin receptor utilizing clatherin-coated endosomes. Mn 3ϩ -transferrin is unlikely to be a predominant manganese species in the brain because rat perfusion studies and neuronal cell uptake assays suggest that the rate of Mn 3ϩ -transferrin uptake is much slower than that of the uptake of Mn 2ϩ species (9,16). This study focuses on the molecular mechanism of Mn 2ϩ uptake by BMVECs. We previously illustrated that DMT1 is expressed by the brain microvasculature but appeared to play little role in the transport of non-transferrin-bound iron (19). DMT1 is a proton/divalent cation transporter, which functions optimally at pH 5.5 (20 -22). On the other hand, the ZIP family proteins, ZIP8 and ZIP14, which have been implicated in manganese uptake (8,(23)(24)(25)(26)(27)(28)(29)(30), function optimally at pH 7.5 and are suggested to be HCO 3 Ϫ /divalent cation symporters (29, 30). We have taken advantage of these functional characteristics to implicate the two ZIP transporters and not DMT1 in Mn 2ϩ uptake into human BMVECs (hBMVECs).
Our work takes advantage of a well-established in vitro blood-brain barrier model composed of an immortalized human-derived cell line of brain microvascular endothelial cells grown in monolayers or polarized in Transwell inserts (19). We followed manganese transport at normal physiologic concentrations (100 -200 nM) and manipulated conditions of manganese acquisition by these endothelial cells to interrogate the protein(s) responsible. Mn 2ϩ uptake is enhanced at basic pH values, by LPS treatment and by increasing HCO 3 Ϫ concentrations, behaviors that support the premise that ZIP8 and/or ZIP14 and not DMT1 function in Mn 2ϩ accumulation by hBMVECs. Treatment with Dynasore, an inhibitor of clatherindependent endocytosis (31,32), Fe 2ϩ , or Mn 2ϩ enhances cellsurface localization of ZIP14 but not ZIP8; the effect of Fe 2ϩ and Mn 2ϩ indicates that ZIP14 recycling at the plasma membrane is regulated by the transporter's ligands. Knockdown of ZIP8 and/or ZIP14 by siRNA reduces Mn 2ϩ accumulation in hBMVECs at both the apical (luminal-facing) and basolateral (abluminal-facing) membranes, indicating that ZIP8 and ZIP14 function in Mn 2ϩ accumulation by hBMVECs at both membranes in these polarized cells. In confirmation of this premise, cell-surface biotinylation experiments in Transwells reveal that ZIP8 and ZIP14 reside on both the apical and basolateral surfaces of hBMVECs and that Mn 2ϩ uptake at those two mem-branes correlates with their occupancy. In summary, these data suggest that ZIP8 and ZIP14 are functional Mn 2ϩ importers at the blood and brain interfaces of the brain's capillary bed.

The kinetic profile of Mn 2؉ accumulation in brain microvascular endothelial cells
The predominant species of manganese in the brain is in the divalent form (9,33). Therefore, we monitored the accumulation of Mn 2ϩ in hBMVECs utilizing 54 Mn 2ϩ as radiotracer with the objective of characterizing the kinetic profiles of uptake and efflux. 54 Mn 2ϩ uptake in hBMVEC monolayers was linear between 10 and 60 min at a rate of 0.125 Ϯ 0.013 pmol/mg/min. From 3 to 24 h, the rate of 54 Mn 2ϩ accumulation was 0.022 Ϯ 0.001 pmol/mg/min (Fig. 1, A and B).
These data suggest that Mn 2ϩ uptake in hBMVECs was biphasic; another interpretation of the data is that the apparent slower accumulation at the later time was due to competing Mn 2ϩ efflux. To test this premise, we investigated the kinetics of Mn 2ϩ efflux. hBMVECs grown in monolayers were incubated with 100 nM 54 Mn 2ϩ for 24 h. After washing, 54 Mn 2ϩ efflux was quantified. Efflux was linear at a rate of 0.049 Ϯ 0.006 pmol/mg/min (Fig. 1C). We conclude that during uptake, cell [Mn 2ϩ ] reaches a level that supports efflux sufficient to reduce the net uptake measured. This interpretation leads to the prediction that Mn 2ϩ accumulation by hBMVECs would reach a steady state. The uptake data shown in Fig. 1B indicate this condition was reached between 24 and 30 h.
To test whether 54 Mn 2ϩ accumulation in hBMVECs was supported by a bulk, endocytic mechanism versus a carriermediated one, we utilized a well-recognized inhibitor of clatherin-coated endocytosis, Dynasore (32,34). In fact, pretreatment with 80 M Dynasore enhanced manganese uptake in hBMVECs compared with the DMSO vehicle (Fig. 1D). This finding is consistent with the premise that by reducing membrane protein retrograde recovery, Dynasore treatment increases the lifetime of a Mn 2ϩ transporter(s) on the plasma membrane.

؊
To help identify this transporter(s), we took advantage of the differential biochemical regulators of the known divalent metal ion transporters (18). Thus, we examined the effects of pH, LPS, and HCO 3 Ϫ on Mn 2ϩ uptake in hBMVECs. As noted, DMT1 functions optimally at pH 5.5, whereas ZIP8 and ZIP14 function best at pH 7.5 (8,22,24,25). Using HEPES buffers (11) ranging from pH 6.2 to 7.8, we demonstrated that 54 Mn 2ϩ accumulation exhibited an ascending pH-rate profile with an increase in uptake of 25-fold over this pH range ( Fig. 2A). In addition, we examined Mn 2ϩ uptake by hBMVECs pretreated with LPS, a pro-inflammatory endotoxin. LPS-treated macrophages and hepatocytes exhibit an increased expression of ZIP8 and ZIP14, respectively (35,36). We found that LPS enhanced Mn 2ϩ accumulation compared with untreated cells (Fig. 2B). Last, the transport function of ZIP8 and ZIP14 depends on the presence of HCO 3 Ϫ (24,25). Thus, we added varying amounts of HCO 3 Ϫ to HBSS uptake buffer and observed a direct relationship between the concentration of extracellular HCO 3 Ϫ and ZIP8 and ZIP14 in brain manganese uptake hBMVEC 54 Mn 2ϩ accumulation (Fig. 2C). Note that in contrast to the 54 Mn 2ϩ uptake in the absence of HCO 3 Ϫ illustrated in Fig.  1A, in the experiment in Fig. 2C, a different loading medium was used. To control for carbonate, HBSS was the loading buffer used in Fig. 2C, whereas RPMI 1640 plus serum growth medium was used in Fig. 1. Collectively, the results presented in Fig. 2 support the conclusion that ZIP8 and/or ZIP14 likely make(s) a majority contribution to Mn 2ϩ accumulation by hBMVECs, whereas DMT1 does not. Based on these findings we sought direct evidence that one or both of these two transporters were primarily responsible for divalent manganese uptake by hBMVECs. In addition, we tested the novel premise that these proteins supported Mn 2ϩ accumulation at both the apical and basal membranes of these polarized endothelial cells.

ZIP8 and ZIP14 are expressed by hBMVECs
Although both ZIP8 and ZIP14 exhibit a broad tissue distribution (30), their expression in hBMVECs has not been reported. To this end, standard ⌬⌬C t data were obtained for both transcripts with respect to ␤-actin ( Fig. 3A) but also with respect to the relative abundance of ZIP8 versus ZIP14 message (Fig. S1). As noted, this latter quantification was based on the

ZIP8 and ZIP14 in brain manganese uptake
amplification efficiency for each set of primers determined from the slope of a C t versus log[cDNA] analysis. The near equivalency of the intercept values of these plots indicate a comparable abundance of ZIP8 and ZIP14 mRNA (37,38). In addition, both proteins were visualized in these cells by indirect immunofluorescence (Fig. 3B). The punctate, cytosolic localization indicated by these images is similar to that reported previously for chromosomally encoded ZIP8 and ZIP14 (24,25). This distribution indicated by immunofluorescence is consistent with the surface biotinylation results presented below.

ZIP14 surface residence is modulated by clatherin-coated endocytosis, LPS, and Mn 2؉
The enhanced Mn 2ϩ accumulation upon Dynasore treatment suggested that it stabilized Mn 2 importer(s) on the surface of hBMVECs. Because treatment of cells with Fe 2ϩ , Mn 2ϩ , or LPS has increased expression of ZIP8 and/or ZIP14 (4,8,30,35,36), we examined whether any of these treatments influenced cell surface ZIP8 or ZIP14 occupancy in hBMVECs.
To quantify such a response, we took a cell-surface biotinylation approach to tag membrane-expressed proteins in hBM-VECs grown in monolayers. After the cells were incubated with the biotin reagent, the cells were lysed; the lysates, normalized by protein content, are presented here as the "input" samples. These samples were fractionated by affinity capture on streptavidin columns. Nonbiotinylated proteins were collected in the flow-through (the "unbound" fractions), whereas tagged, cellsurface proteins were eluted off the streptavidin column by reductive cleavage of the disulfide bond linker; this is the "bound" fraction. The abundance of biotinylated, bound protein species was normalized based on the protein content of the input. These fractions were analyzed by Western blotting to establish whether Dynasore, Fe 2ϩ , Mn 2ϩ , or LPS treatments affected ZIP8 and ZIP14 surface localization ( Fig. 4 and Figs. S2 and S3).
ZIP8 and ZIP14 band intensities in all of the fractions were normalized to ␤-actin in the input, appropriate because all samples processed were adjusted to maintain constant total protein concentration. The data show that none of these treatments had a statistically significant effect on the surface presentation of ZIP8. In contrast, Mn 2ϩ , Dynasore, and LPS increased ZIP14 surface residence 2-fold compared with untreated cells without, however, any effect on the total cell ZIP14 protein content ( Fig. 4A and Figs. S2 and S3). These effects are quantified in Fig.  4 (B-D). Note that the band intensities in the bound fractions ( Fig. S2 and S3) are stronger than in the input because of their concentration in the affinity capture protocol and a difference in exposure time. However, by normalizing input and bound protein abundance based on these factors, the fraction of total cell transporter in the membrane was estimated. This estimate for both ZIP8 and ZIP14 was 5%, a value consistent with the proteins' primarily intracellular localization indicated by the images presented in Fig. 3B.

ZIP8 and ZIP14 facilitate Mn 2؉ accumulation in hBMVECs
The dependence of Mn 2ϩ uptake on pH, for example, and the cell-surface localization of ZIP8 and ZIP14 support the inference that these transporters are the primary Mn 2ϩ import proteins in hBMVECs. To test this hypothesis, siRNA delivery via electroporation was used to knock down endogenous ZIP8 and/or ZIP14 transcripts. With this approach, SLC39A8 mRNA was reduced by 85%, and SLC39A14 transcript was reduced by 65% in the single knockdowns, and each were reduced by 50% in the double knockdown (DKD) samples (Fig. 5, A and B, respectfully). siRNA knockdown of SLC39A8 decreased ZIP8 protein by 25%, and siRNA knockdown of SLC39A14 reduced ZIP14 protein by 50% at 2 days post-transfection ( We used these knockdown conditions to examine the effect on 54 Mn 2ϩ accumulation. Under 200 nM 54 Mn 2ϩ -loading conditions, single knockdowns of ZIP8 and ZIP14 reduced Mn 2ϩ uptake in hBMVECs ϳ50% compared with the siGLO negative controls (Fig. 5E). In the DKD conditions, there was a synergistic drop in 54 Mn 2ϩ accumulation (Fig. 5E). These data indicate that both proteins support Mn 2ϩ uptake by hBMVECs.
Our next objective was to determine how knockdown of these transporters altered uptake at [Mn] Ͼ35 g/liter, a blood marker of manganese overload; we used 1 M 54 Mn 2ϩ to reproduce this condition. Knockdown of either ZIP8 and ZIP14 reduced 54 Mn 2ϩ uptake by 20%, which was further decreased to 50% in the DKD samples compared with the siGLO controls (Fig. 5F). Although the differences were not as statistically robust as those observed under conditions of physiological Mn 2ϩ loading, ZIP8 and ZIP14 appear, nonetheless, to support Mn 2ϩ accumulation under manganese-overload conditions as well. Although there was a consistent correlation between the percentage of knockdown of transporter abundance and a decrease in Mn 2ϩ uptake, our data do not strictly exclude the possibility that another divalent metal ion transporter contributes to the 54 Mn 2ϩ accumulation quantified. Transcript abundance of each protein was quantified by qPCR (A). RNA was collected, reverse-transcribed and resulting cDNA was quantified by qPCR. The relative fold expression of ZIP8 and ZIP14 presented was calculated by the standard ⌬⌬C t method. ZIP8 and ZIP14 in hBMVECs were visualized by indirect immunofluorescence (B). The cells were grown on sterile coverslips to 85% confluency, fixed, blocked, and incubated overnight with primary antibodies to ZIP8 (top left panel) or ZIP14 (bottom left panel). Coverslips were incubated in anti-rabbit Alexa Fluor 647-conjugated secondary for 1 h followed by a 10-min nuclear stain with Hoescht 33342. Coverslips were mounted using Prolong Gold mounting media and sealed. Images were acquired at 63ϫ magnification with oil immersion on a Leica TCS SP8 confocal microscope. Negative controls (right panel) were obtained by omitting the primary antibody incubation.

ZIP8 and ZIP14 in brain manganese uptake ZIP8 and ZIP14 function at both the apical and basolateral membranes of hBMVECs
hBMVECs are polarized cells with luminal (or apical; bloodfacing) and abluminal (or basolateral; brain-facing) membranes (39). Mn 2ϩ from the circulation crosses the apical membrane to supply the brain with a dietary manganese source via efflux from the basolateral membrane (40). Thus, we examined whether ZIP8 and ZIP14 are localized on the apical membrane of hBMVECs. We seeded hBMVECs in Transwells to establish an in vitro BBB model and quantified 54 Mn 2ϩ accumulation from the apical chamber. As a control, we compared apical to basolateral Mn 2ϩ uptake. Unexpectedly, uptake at the basolateral membrane was more robust than at the apical membrane (Fig. 6, compare B with A). This differential in flux trajectory is illustrated in Fig. 6C; the more robust uptake at the basal surface is indicated also by the more robust 54 Mn 2ϩ accumulation by hBMVECs when loaded from the basal compartment (Fig.  6D). Note that in addition to using transendothelial electrical resistance (TEER) measurement to ensure retention of barrier integrity in these experiments, this differential flux trajectory by itself also demonstrates the absence of paracellular leakage.
The novel finding that hBMVECs accumulate Mn 2ϩ from the abluminal space as well as from the "blood" indicate that ZIP8 and ZIP14 are localized and function as Mn 2ϩ importers on both apical and basolateral membranes. We assessed this by using siRNA to knock down ZIP8 and/or ZIP14. hBMVECs were grown in Transwell inserts, and Mn 2ϩ uptake was quantified from both apical and basal chambers. Knockdown of ZIP8 statistically decreased apical Mn 2ϩ uptake by ϳ50% compared with the siGLO controls (Fig. 6E). Although not significant (p ϭ 0.147), knockdown of ZIP14 reduced apical Mn 2ϩ uptake by ϳ40% and the DKD samples reduced uptake by ϳ85% compared with the control. When the cells were loaded from the basal chamber, knockdown of ZIP8 reduced Mn 2ϩ accumulation by 50%, whereas ZIP14 knockdown diminished uptake by ϳ25%, and the DKD lowered Mn 2ϩ uptake by 60% compared with the siGLO controls (Fig. 6F). These results indicate that both ZIP8 and ZIP14 are apical and basolateral Mn 2ϩ importers in hBMVECs.
Importantly, we confirmed that ZIP8 and ZIP14 localize to both membranes of hBMVECs. To quantify this distribution, we again used cell-surface biotinylation of cells grown in Tran- The arrowheads denote the monomeric forms of the ZIP proteins; * denotes oligomeric forms. Quantification of ZIP8 and ZIP14 surface fractionation compared with untreated cells, normalized to ␤-actin (B). Statistically significant difference was calculated using Tukey's multiple comparison test, with n ϭ 3 experimental replicates/condition. *, p Ͻ 0.05; ***, p Ͻ 0.001. n ϭ 3 experimental replicates/condition. hBMVECs were treated with 1 g/ml LPS for 3 or 24 h prior to cell surface biotinylation (C and D). Input and bound fractions were processed for Western blotting (see Fig. S3 to see blots). Band intensities for ZIP8 (C) and ZIP14 (D) expression were normalized to ␤-actin. Statistically significant difference was measured by Tukey's multiple comparison test. *, statistically significantly different compared with untreated cells. *, p Ͻ 0.05, with n ϭ 3 or more replicates/condition. ZIP8 and ZIP14 in brain manganese uptake swells. Membrane-specific labeling is possible in this paradigm via the addition of reagent to either the apical or basal chamber. Following pulldown of biotinylated proteins and Western blotting analysis, we confirmed that ZIP8 and ZIP14 are localized to both membranes (Fig. 7, A and B, respectively). Although not statistically significant at p Ͻ 0.1, we observed the fractional distribution of surface ZIP8 to be more apically localized (58%; Fig. 7A), whereas surface ZIP14 was preferentially localized to the basolateral membrane (91%; Fig. 7B).

Discussion
The objective of this study was to elucidate the transporter(s) responsible for Mn 2ϩ uptake in hBMVECs. Inasmuch as these cells support the barrier that filters metabolites in the general circulation for their permeation into the abluminal space, we chose them as an appropriate model for brain manganese trafficking. Clinical presentation of patients carrying mutations in either ZIP8 or ZIP14 has implicated these two promiscuous divalent metal ion transporters in brain manganese metabolism (4,8,27,(41)(42)(43)(44)(45)(46)(47)(48). Our data demonstrate at the protein and cell level that both contribute to the trafficking of Mn 2ϩ in the endothelial cells found in the brain's capillary bed.
We examined the kinetics of manganese uptake and efflux in hBMVECs using 54 Mn 2ϩ , the likely ionic species found in the circulation, and at a concentration of 100 -200 nM that is within the physiological range of circulating manganese. Our data show that at steady state, Mn 2ϩ uptake and efflux over time results in a net 54 Mn 2ϩ accumulation by hBMVECs. We considered the possibility that bulk transport (i.e. endocytosis) contributed to Mn 2ϩ uptake. However, Dynasore, a clatherincoated endocytosis inhibitor, induced an increase, not a decrease, in Mn 2ϩ uptake by hBMVECs. Inhibiting caveolinmediated endocytosis with genistein also enhanced 54 Mn 2ϩ accumulation (data not shown). These data indicate that bulk vesicular transport does not support Mn 2ϩ accumulation but are consistent with the model that inhibition of protein turnover/recycling at the plasma membrane amplifies Mn 2ϩ uptake.
The candidates for Mn 2ϩ import are ZIP8, ZIP14, and DMT1; we show here and in previous work from this lab (19) that all three divalent metal ion transporters are expressed by hBMVECs. To determine which of the three is/are involved in Mn 2ϩ uptake by hBMVECs, we investigated the influence of varying pH in a 54 Mn 2ϩ accumulation assay. Mn 2ϩ uptake in hBMVECs was strongly responsive to increasing pH, indicating that ZIP8 and/or ZIP14 and not DMT1 are the primary divalent metal ion transporters in hBMVECs. Previous work from this lab showed that non-transferrin-bound iron uptake in hBMVECs also utilized a DMT1-independent uptake mechanism (19). We demonstrated here also that LPS and HCO 3 Ϫ increased Mn 2ϩ accumulation in hBMVECs, consistent with  A and B). RNA was collected 2 days postelectroporation and reverse-transcribed, and the resulting cDNA was quantified by qPCR. The samples were compared with siGLO controls and normalized to ␤-actin transcript. # , significantly different compared with siGLO control; ### , p Ͻ 0.001; ***, p Ͻ 0.001. n ϭ 9 or more /siRNA condition. ZIP8 and ZIP14 protein expression is reduced in siRNA knockdown samples (C and D). Lysates of siRNA samples were collected 2 days postelectroporation, cell surface-biotinylated, and processed for Western blotting. The samples were compared with siGLO controls and normalized to ␤-actin transcript levels. *, significantly different compared with siGLO control. *, p Ͻ 0.05; **, p Ͻ 0.01. n ϭ 3-4 samples/siRNA condition. Physiologic 54 Mn 2ϩ accumulation is reduced in ZIP8 and ZIP14 knockdown hBMVECs (E). hBMVECs were loaded with 54 Mn 2ϩ (200 nM) for 1 h with uptake normalized for protein concentration. Statistical significance was tested by Tukey's multiple comparison test. # , significantly different compared with siGLO control; ### , p Ͻ 0.001; **, p Ͻ 0.01; ***, p Ͻ 0.001. n ϭ 6 experimental replicates/sample. 54 Mn 2ϩ accumulation is somewhat reduced in ZIP8 and ZIP14 knockdown hBMVECs challenged with a hypernormal [Mn] (F). hBMVECs were loaded with 54 Mn 2ϩ (1 M) for 1 h with uptake normalized for protein concentration. Statistical significance was tested by Tukey's multiple comparison test. # , significantly different compared with siGLO control; ### , p Ͻ 0.001; **, p Ͻ 0.01. n ϭ 6 experimental replicates/sample.

ZIP8 and ZIP14 in brain manganese uptake the known effects of LPS and HCO 3
Ϫ on ZIP8 and ZIP14 expression and activity (24,25,35).
As noted, we concluded that Dynasore enhances Mn 2ϩ accumulation by blocking surface protein internalization. Within the L2 cytosolic loop of ZIP14, there is a YSDI motif suggested to be a tyrosine-based endocytosis signal (49,50). The YTRF sequence in the transferrin receptor is an example of such a motif (51). In addition, Zhao et al. (52) reported that ZIP14 surface residence was increased in HepG2 cells when treated with Dynasore. Our results in hBMVECs parallel this behavior as treatment with Dynasore increased the membrane expres-sion of ZIP14; in our work, an added finding was that such treatment had no effect on ZIP8 localization. In addition, knockdown of ZIP14 treated with Dynasore resulted in reduced 54 Mn 2ϩ uptake compared with siGLO and siZIP8 Dynasoretreated cells, consistent with the model that Mn 2ϩ uptake correlates in part with the membrane presentation of ZIP14, specifically (Fig. S5).
Fe 2ϩ , Mn 2ϩ , and LPS have been reported to modulate expression of ZIP8 and ZIP14 (4,5,30,35,36). We were interested in how these treatments affected ZIP8 and ZIP14 expression and plasma membrane residence in hBMVECs. The results show that Fe 2ϩ , Mn 2ϩ , and LPS treatments do not alter the abundance of ZIP8 or ZIP14, whereas Mn 2ϩ and LPS promote increased ZIP14 surface occupancy 2-fold. No significant change in surface expression of ZIP8 was observed following either of these treatments. These results suggest also the model that cycling of ZIP14 to and from the plasma membrane is a contributor to the regulation of Mn 2ϩ accumulation by hBMVECs.
Knockdown of ZIP8 and ZIP14 mRNA was robust but did not result in a complete knockdown of either protein. Consequently, we did not observe a complete inhibition of Mn 2ϩ uptake under any knockdown condition. ZIP8 knockdown resulted in a 25% reduction of the protein, whereas the ZIP14 knockdown reduced protein expression by 50%. Despite this dissimilar knockdown efficiency, there was no statistically significant difference in the loss of 54 Mn 2ϩ accumulation efficiency. Interestingly, whereas knockdown of SLC39A14 upregulated SLC39A8 transcript, there was no difference in SLC39A14 expression following knockdown of SLC39A8 (Fig.  S6). Other data have shown that SLC39A8 was able to compensate for the lack of SLC39A14 (23,53,54), results that are similar to those we report here for hBMVECs. However, the importance of both proteins in Mn 2ϩ accumulation is indicated by the fact that the knockdown of both transporters reduces 54 Mn 2ϩ uptake more than the single knockdowns. On the other hand, when challenged with nonphysiologic [Mn 2ϩ ], the knockdown of both transporters had a more limited effect, suggesting that additional uptake mechanisms obtain under conditions of systemic manganese overload.
The restrictive nature of the BBB is due to the tight junctions between BMVECs caused by expression of claudins and occludins, which inhibit paracellular metabolite trafficking (39). Under normal conditions, nutrients and solutes enter the brain solely via transcellular transport. To delineate the mechanism of Mn 2ϩ handling at this barrier in vitro, we grew hBMVECs in Transwell inserts, allowed the cells to polarize, and compared apical versus basolateral Mn 2ϩ accumulation. Our Transwell model allowed us to determine whether ZIP8 and ZIP14 were localized to both the apical and basolateral membranes. Our surface biotinylation protocol demonstrated the expression of both proteins in both membranes; furthermore, we demonstrated that ZIP14 was the predominant transporter at the basal surface. Our knockdown of the expression of the two transporters in hBMVECs grown in Transwells and subsequent quantification of 54 Mn 2ϩ uptake was consistent with this differential distribution in that ZIP14 knockdown had a more robust effect on basal compared with apical Mn 2ϩ uptake.  54 Mn 2ϩ in each chamber was quantified over time. Statistical significance of change was tested by Tukey's multiple comparison test. *, statistically significantly different from 0.5 h; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ∧, difference from 1 h; ∧, p Ͻ 0.05; ∧∧∧, p Ͻ 0.001. # , difference from 3 h; # , p Ͻ 0.05; ## , p Ͻ 0.01; ### , p Ͻ 0.001. ϩ , difference from 6 h; ϩϩϩ , p Ͻ 0.001. Based on these data the relative 54 Mn 2ϩ flux trajectory was calculated (C). After 24 h, the cells were lysed, and 54 Mn 2ϩ accumulation was normalized for protein concentration (D). Statistically significant difference was tested by unpaired t test. *, p Ͻ 0.05; ****, p Ͻ 0.0001. hBMVECs were electroporated with siRNAs for ZIP8 and ZIP14 and grown in the apical chamber of Transwell Thincerts TM (E and F). The cells were loaded with 100 nM 54 Mn 2ϩ for 3 h from the apical (E) or basal chamber (F). Statistically significant difference was measured using Tukey's multiple comparison test. *, statistically significantly different compared with siGLO; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ∧∧, p Ͻ 0.01. n ϭ 4 experimental replicates/condition.

ZIP8 and ZIP14 in brain manganese uptake
The work here in the barrier cells of the brain is consistent with the recent report that in the intestinal epithelial Caco2 cell line grown in Transwells, ZIP14 is expressed on the basal (blood) membrane and supports a basolateral-to-apical (secretory) manganese transport (55). The pattern that appears to be developing is that at such barriers, Mn 2ϩ transport is bidirectional. This transcellular trafficking has not been considered as part of the systemic handling of this transition metal including at the BBB.
With this model in mind, the effects on manganese distribution in KO mouse lines are instructive. Inducible ZIP8 knockout mice display decreased blood and brain manganese levels compared with control mice (27). Our cell-surface biotinylation and functional ZIP8 knockdown studies directly correlate with the decrease in manganese accumulation in the brains of these mice. In contrast, whole-body ZIP14 knockout mice have elevated blood and brain manganese levels (23,26,54). Patients with genetic mutations in SLC39A14 also exhibit brain manganese overload (8,43). The elevated blood manganese-level in the KO mice is fully consistent with the basal-to-apical, secretory manganese trafficking in Caco2 cells recently reported (55). Our data provide a cellular mechanism for the brain manganese accumulation quantified in the mouse model and associated with SLC39A14-linked genetic disorders. Our quantitative cell-surface biotinylation assay in Transwells showed that ZIP14 mainly resides in the basolateral membrane. This is in agreement with studies reporting basolateral localization of ZIP14 in hepatocytes, enterocytes, and Caco-2 cells, as noted (28, 56 -58). Thus, with this view of ZIP14 function, whereas ZIP8 continues to support hBMVEC Mn 2ϩ uptake and brain manganese accumulation, the secretory flow of Mn 2ϩ out of the brain normally supported by ZIP14 is absent. The result is brain Mn 2ϩ accumulation. Clearly, our simple in vitro model does not reflect the complexity inherent in examination of whole-body manganese distribution, but it does replicate the characteristics of the in vivo clinical and mouse model outcomes.
In summary, we leveraged the biochemical behaviors of the putative mammalian Mn 2ϩ transporters and siRNA knockdown to demonstrate that ZIP8 and ZIP14 are the dominant proteins associated with manganese uptake in hBMVECs. Additionally, we show that Mn 2ϩ and LPS modulate ZIP14 but not ZIP8 membrane recycling. Importantly, we provide evidence for ZIP8-and ZIP14-dependent Mn 2ϩ uptake by hBMVECs at both the apical and basolateral membranes of these cells. This novel finding indicates that these two transporters support a bidirectional, transcellular flux of Mn 2ϩ at the blood-brain barrier. This flux, of course, requires a corresponding efflux pathway whose characteristics are under current delineation.

Cell culture
hBMVECs were a generous gift from Dr. Supriya Mahajan (University at Buffalo); the generation and characteristics of this cell line have been described in detail (59,60) and have been validated as an hBMVEC cell line (59). hBMVECs were cultured as previously described (19), reaching ϳ90 -95% confluency at the time of the experiment. The experiments were performed in 3-cm, 6-cm, or 24-well tissue culture dishes or Greiner Bioone Transwells (1.0-m pore size), as specified. For hBMVECs grown in Transwells, barrier integrity was measured by TEER quantified using a Millicell ERS-2 Epithelial Volt-Ohm meter according to the manufacturer's instructions (EMD Millipore, Billerica, MA).

Mn 2؉ accumulation and efflux assays
For 54 Mn 2ϩ accumulation assays, hBMVEC monolayers were loaded with physiological concentrations (100 nM) of was quantified as a percentage of the total of each protein found to be membrane-associated. Statistically significant difference was analyzed using unpaired t test. *, p Ͻ 0.05. Note that the actin control is the same in A and B because the same blot was used to probe for ZIP8 and ZIP14 in the input fractions to which the bound fractions were normalized.
ZIP8 and ZIP14 in brain manganese uptake 54 MnCl 2 (PerkinElmer), in RPMI 1640 plus serum growth medium, unless otherwise indicated. hBMVECs grown in Transwell inserts were loaded with 100 nM (or 200 nM for the flux assay) 54 Mn 2ϩ in the apical chamber (RPMI 1640 plus serum growth medium) or in the basal chamber (RPMI 1640 minus serum, plus 5 g/ml human insulin and 30 nM sodium selenite growth medium). The reactions were terminated with ice-cold quench buffer as previously described (61) and then lysed with RIPA buffer (25 mM Tris, 150 mM NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS).
For 54 Mn 2ϩ efflux assays, hBMVEC monolayers were loaded with 100 nM 54 Mn 2ϩ for 24 h. The cells were washed once with RPMI 1640 (minus serum, 5 g/ml human insulin, 30 nM sodium selenite) and twice with 250 M sodium citrate. 54 Mn 2ϩ efflux was monitored in RPMI 1640 plus serum growth medium for up to 3 h. Efflux was quenched as above, and the cells lysed with RIPA buffer. The lysates were analyzed for 54 Mn 2ϩ and protein content. Cell-associated 54 Mn 2ϩ counts (LKB Wallac CompuGamma) were normalized to protein concentration. Protein was quantified by BCA assay (Thermo Scientific).

siRNA delivery by electroporation
SLC39A8 and SLC39A14 expression was knocked down in hBMVECs using siGENOME TM human SLC39A8 siRNA SMARTpool and siGENOME TM human SLC39A14 siRNA SMARTpool, respectively (Dharmacon). RISC-independent siGLO transfection indicator (Dharmacon) was used as a negative control to infer transfection efficiency. The siGLO control contains a fluorescein amidite-labeled reporter to confirm appropriate nuclear localization of the siRNAs. The siRNAs were delivered via electroporation using the NEON transfection system (Life Technologies). Electroporation was conducted with two pulses of 1150 V for 30 ms. hBMVECs were electroporated with 125 nM siRNAs (final concentration in the culture dish) in RPMI 1640 growth medium with 10% FBS (no antibiotics). The medium was changed 24 h after electroporation, and cell growth, viability, and transfection efficiency (siGLO) were noted. All assays using siRNA knockdown samples were performed 48 h postelectroporation.

RT-qPCR
Total RNA was extracted from hBMVEC cells using the TRIzol reagent (Invitrogen) as per the manufacturer's instructions. After DNase treatment and purification with Direct-zol RNA kit (Zymo Research), 400 ng of RNA was reverse-transcribed using the iScript TM cDNA synthesis kit (Bio-Rad) to generate cDNA. Real-time quantitative PCR was performed using SsoAdvanced TM Universal SYBR Green Supermix and analyzed using CFX96 Touch TM real-time PCR detection system (Bio-Rad). Relative expression of the target gene was normalized to ␤-actin and calculated using the ⌬⌬C t method. End point qPCRs were separated on a 1.7% agarose gel to confirm product size. The primer sequences used to amplify for each transcript were as follows: human SLC39A8 forward, 5Ј-ATG-CTACCCAAATAACCAGC-3Ј; human SLC39A8 reverse, 5Ј-CAGGAATCCATATCCCCAAAC-3Ј; human SLC39A14 forward, 5Ј-TAAGCAGAAAAATGAGCATC-3Ј; human SLC39A14 reverse, 5Ј-ACCTTTCAGCCAGTAGCAAG-3Ј; ␤-actin forward, 5Ј-GGGTCAGAAGGACTCCTACG-3Ј; and ␤-actin reverse, 5Ј-GGTCTCAAACATGATCTGGG-3Ј.

Cell-surface biotinylation
Membrane proteins were separated from hBMVECs as previously described (62). Briefly, cells grown in 6-well plates or 6-well Transwell inserts were washed twice with PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 (used for entire biotinylation protocol) and then treated with 0.5 mg/ml EZ-Link TM Sulfo-NHS-SS-Biotin (Thermo Fisher) for 30 min (or 2 h for Transwells) at 4°C. Then the cells were washed twice with PBS (ϩCa 2ϩ , Mg 2ϩ ) containing 0.1% BSA and twice with PBS (ϩCa 2ϩ , Mg 2ϩ ). The cells were lysed by scraping in ice-cold RIPA buffer (25 mM Tris, 150 mM NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS, pH 7.4) supplemented with 4ϫ Halt protease inhibitor mixture (Thermo Fisher) and incubation on ice for 15 min. The cell suspension was then centrifuged at 10,000 ϫ g for 10 min at 4°C, and the supernatant (input fraction) was collected and loaded onto a NeutrAvidinagarose column overnight at 4°C. Following collection of the flow-through (unbound fraction), the columns were washed with RIPA buffer containing 4ϫ protease inhibitors and eluted at 50°C in 6ϫ SDS-loading buffer, 150 mM DTT (bound fraction). The input and unbound fraction protein content was quantified, and equal amounts of protein were loaded for Western blotting. All of the elution from the bound fraction was used for SDS-PAGE followed by Western blotting.

Indirect immunofluorescence
hBMVECs grown on coverslips were fixed for 10 min with 3.7% paraformaldehyde and 4% sucrose in PBS (ϩCa 2ϩ , Mg 2ϩ ), washed twice with PBS, and incubated for 1 h in 1% BSA, 0.3 M ZIP8 and ZIP14 in brain manganese uptake glycine, 0.1% Tween 20. The cells were incubated overnight at 4°C with anti-ZIP8 or anti-ZIP14 antibodies (1:1000 dilution) in 1% BSA. The cells were washed at room temperature for 5 min three times and then probed with 1:1000 secondary donkey anti-rabbit Alexa 488-conjugated goat anti-rabbit (A21206; Invitrogen) in PBS 1% BSA for 1 h at room temperature followed by three 5-min washes (all washes were with PBS containing Ca 2ϩ and Mg 2ϩ as above). The cells were stained with 0.7 g/ml Hoechst 33342 (Thermo Fisher Scientific) for 10 min at room temperature, followed by three 5-min washes. The coverslips were mounted onto glass microscope slides using Prolong TM glass mounting medium (Invitrogen). Images were obtained using a Leica TCS SP8 confocal microscope and were analyzed by Vaa3D software. As noted in the figure legend, the images were collected at 63-magnification.

Statistical data analysis
All statistical analyses were performed using Prism 5.0 or 8.0 (GraphPad Software, San Diego, CA). The data are presented as means Ϯ S.D. Unpaired t tests were used when comparisons were made between two conditions (one variable) from the same time point. Comparisons of multiple samples were made using one-way analysis of variance statistical analyses in conjunction with Tukey's multiple comparison tests. In the text, "n" refers to the number of technical replicates used to derive the stated quantity.