The intestinal metal transporter ZIP14 maintains systemic manganese homeostasis

ZIP14 (encoded by the solute carrier 39 family member 14 (SLC39A14) gene) is a manganese transporter that is abundantly expressed in the liver and small intestine. Loss-of-function mutations in SLC39A14 cause severe hypermanganesemia. Because the liver is regarded as the main regulatory organ involved in manganese homeostasis, impaired hepatic manganese uptake for subsequent biliary excretion has been proposed as the underlying disease mechanism. However, liver-specific Zip14 KO mice exhibit decreased manganese only in the liver and do not develop manganese accumulation in other tissues under normal conditions. This suggests that impaired hepatobiliary excretion is not the primary cause for manganese overload observed in individuals lacking functional ZIP14. We therefore hypothesized that increased intestinal manganese absorption could induce manganese hyperaccumulation when ZIP14 is inactivated. To elucidate the role of ZIP14 in manganese absorption, here we used CaCo-2 Transwell cultures as a model system for intestinal epithelia. The generation of a ZIP14-deficient CaCo-2 cell line enabled the identification of ZIP14 as the major transporter mediating basolateral manganese uptake in enterocytes. Lack of ZIP14 severely impaired basolateral-to-apical (secretory) manganese transport and strongly enhanced manganese transport in the apical-to-basolateral (absorptive) direction. Mechanistic studies provided evidence that ZIP14 restricts manganese transport in the absorptive direction via direct basolateral reuptake of freshly absorbed manganese. In support of such function of intestinal ZIP14 in vivo, manganese levels in the livers and brains of intestine-specific Zip14 KO mice were significantly elevated. Our findings highlight the importance of intestinal ZIP14 in regulating systemic manganese homeostasis.

ZIP14 (encoded by the solute carrier 39 family member 14 (SLC39A14) gene) is a manganese transporter that is abundantly expressed in the liver and small intestine. Loss-of-function mutations in SLC39A14 cause severe hypermanganesemia. Because the liver is regarded as the main regulatory organ involved in manganese homeostasis, impaired hepatic manganese uptake for subsequent biliary excretion has been proposed as the underlying disease mechanism. However, liver-specific Zip14 KO mice exhibit decreased manganese only in the liver and do not develop manganese accumulation in other tissues under normal conditions. This suggests that impaired hepatobiliary excretion is not the primary cause for manganese overload observed in individuals lacking functional ZIP14. We therefore hypothesized that increased intestinal manganese absorption could induce manganese hyperaccumulation when ZIP14 is inactivated. To elucidate the role of ZIP14 in manganese absorption, here we used CaCo-2 Transwell cultures as a model system for intestinal epithelia. The generation of a ZIP14-deficient CaCo-2 cell line enabled the identification of ZIP14 as the major transporter mediating basolateral manganese uptake in enterocytes. Lack of ZIP14 severely impaired basolateral-to-apical (secretory) manganese transport and strongly enhanced manganese transport in the apical-to-basolateral (absorptive) direction. Mechanistic studies provided evidence that ZIP14 restricts manganese transport in the absorptive direction via direct basolateral reuptake of freshly absorbed manganese. In support of such function of intestinal ZIP14 in vivo, manganese levels in the livers and brains of intestine-specific Zip14 KO mice were significantly elevated. Our findings highlight the importance of intestinal ZIP14 in regulating systemic manganese homeostasis.
Manganese is an essential nutrient. As a cofactor for several enzymes, it is required for the normal function of several physiologic processes including protein glycosylation, detoxification of superoxide and ammonia, and gluconeogenesis (1,2). Excessive manganese accumulation, however, leads to mangan-ism, whose symptoms resemble those of Parkinson's disease (1). Therefore, manganese homeostasis must be tightly controlled at the systemic and cellular levels.
Systemic manganese homeostasis is maintained by intestinal manganese absorption and hepatobiliary manganese excretion (3)(4)(5)(6). Oral exposure is the major source of manganese absorption with ϳ2-3% of manganese being absorbed from a diet adequate in manganese (7). For dietary manganese intakes that are in the physiologic range, oral manganese absorption appears to be primarily controlled by varying intestinal absorption rates (8). However, biliary manganese elimination becomes the main homeostatic mechanism when intestinal control mechanisms are overwhelmed or bypassed, e.g. by intravenous injection or inhalation exposure.
On a cellular level, the distribution of manganese involves mechanisms for entry and exit, as well as specific pathways that ensure intracellular transport and targeting. To date, these mechanisms remain poorly defined. Components of the iron metabolic pathways; citrate, choline, dopamine, and calcium transporters; the sodium-calcium exchanger; as well as the ATPases Park9 and SPCA1 have been implicated in manganese metabolism (9). More recently, spurred by the discovery of human mutations, members of the ZnT and ZIP families of metal transporters, namely ZnT10, ZIP8, and ZIP14, have been identified as crucial players in cellular manganese metabolism in vivo (10 -14).
Previous studies have shown that ZIP14 is capable of mediating the uptake of manganese in vitro (15). Highlighting the function of ZIP14 as a manganese transporter in vivo, homozygous mutations of ZIP14 result in manganese hyperaccumulation in the blood and other tissues of affected individuals without apparent effects on other metals (10, 16 -18). Initially it was postulated that the primary function of ZIP14 is hepatic manganese uptake for subsequent biliary excretion and that the manganese hyperaccumulation observed in other tissues is secondary to the defect of hepatic ZIP14 (10). The latter hypothesis, however, was falsified by experiments involving a liver-specific Zip14 KO mouse line (19). Specifically, these mice did not display elevated manganese in the blood or other tissues in contrast to global Zip14 KO mouse models, although ZIP14 has been clearly demonstrated to be required for hepatobiliary manganese excretion under high manganese conditions (19 -22).
ZIP14 is also highly abundant in the small intestines of humans and mice (23,24). We hypothesize that, if not impaired hepatobiliary excretion, increased intestinal man-ganese absorption could be the primary cause for the manganese hyperaccumulation observed in individuals lacking functional ZIP14. To test our hypothesis, we generated a CaCo-2 cell line with ZIP14 inactivation (CaCo-2 ⌬ZIP14 cells). Differentiated monolayers of this cell line exhibited a dramatic defect in basolateral manganese uptake. Apical-tobasolateral manganese transport in CaCo-2 ⌬ZIP14 cells was strongly increased compared with that of WT CaCo-2 (CaCo-2 WT ) cells. Further mechanistic studies provided evidence that ZIP14 controls enterocyte manganese absorption by mediating basolateral reuptake of freshly absorbed manganese. To evaluate the physiologic relevance of this novel homeostatic mechanism, we generated intestine-specific Zip14 KO mice. These mice developed markedly increased manganese levels in their livers and brains. Overall, our results suggest an important function of intestinal ZIP14 in the control of systemic manganese homeostasis.

Generation of a ZIP14 KO CaCo-2 cell line
To assess the function of ZIP14 in enterocytes, we inactivated the ZIP14 gene in CaCo-2 cells via CRISPR/Cas9-medi-ated 2 gene KO by targeting the start codon in exon 2 of ZIP14 (Fig. 1A). CaCo-2 cells differentiate spontaneously into polarized enterocytes when grown on permeable filter supports (25). Such CaCo-2 cell cultures provide a well-established model system to study the uptake as well as absorptive (apical-to-basolateral direction) and secretory (basolateral-to-apical direction) fluxes of drugs and nutrients in the intestinal epithelium. This model has been extensively used to study the mechanisms of intestinal iron uptake and transport. In general, results from iron absorption studies conducted in vitro using CaCo-2 cells correlate well with observations made in vivo (26), validating the use of the CaCo-2 cell model for studying intestinal metal uptake and transport.
By using the CRISPR/Cas9 approach, we isolated a CaCo-2 cell clone termed CaCo-2 ⌬ZIP14 with a single-base pair inser-  A, schematic illustration of CRISPR/Cas9-mediated genome editing to generate CaCo-2 ⌬ZIP14 cells. The DNA sequence covering the start codon of ZIP14 within exon 2 is shown. The guide RNA sequence is marked in red letters, and the PAM (protospacer adjacent motif) sequence is underlined in blue. The red arrow marks the predicted Cas9 cleavage site. B, chromatograms of sequencing results using DNA isolated from CaCo-2 WT and CaCo-2 ⌬ZIP14 cells. The black arrow indicates the insertion of an adenosine at the predicted Cas9 cleavage site. C, the single adenosine insertion results in a frameshift with premature translation termination after amino acid 26. D, ZIP14 immunoblot of whole-cell lysates of CaCo-2 WT (WT) and CaCo-2 ⌬ZIP14 (⌬ZIP14) cells confirms the deletion of ZIP14 in CaCo-2 ⌬ZIP14 cells. Both ␤-ACTIN and GAPDH were used as loading controls.

Characterization of CaCo-2 WT and CaCo-2 ⌬ZIP14 Transwell cultures
To study manganese transport across the intestinal epithelium and the role of ZIP14 in this process, both CaCo-2 WT and CaCo-2 ⌬ZIP14 cells were grown into differentiated monolayers for at least 21 days. The presence of polarized monolayers was confirmed by confocal microscopy using the tight junction protein ZO-1 as the apical marker and Na ϩ ,K ϩ -ATPase as the basolateral marker ( Fig. 2A). Further analysis of these images revealed diameters of 11 and 9 m and heights of 12 and 19 m for CaCo-2 WT cells and CaCo-2 ⌬ZIP14 cells, respectively. The consequential larger basolateral and apical surface areas and volumes of CaCo-2 ⌬ZIP14 Transwell cultures, which we estimated to be ϳ180 and 150% of equivalent CaCo-2 WT cultures, are mirrored by a similar increase in the protein contents of these cultures (Fig. 2B). Thus, to compensate for the increased surface, we normalized our uptake and transport data to the protein content in the respective Transwell insert. The integrity of the monolayers was tested by the Lucifer yellow (LY) rejection assay. Apparent LY permeability (P app ) coefficients below 10 nm ϫ s Ϫ1 have been reported to be indicative of well-established CaCo-2 monolayers (27). P app coefficients determined for both CaCo-2 WT and CaCo-2 ⌬ZIP14 monolayers fell in that range but were significantly higher in CaCo-2 ⌬ZIP14 Transwell cultures (Fig. 2C). Because of the smaller diameter of CaCo-2 ⌬ZIP14 cells, CaCo-2 ⌬ZIP14 Transwell cultures possess a greater number of paracellular pores, the sites of paracellular transport, which may partially account for the increased permeability of LY. Consequently, when normalized to protein, P app did not differ significantly between the cultures.
Deletion of Zip14 in mice does not affect the intestinal contents of manganese, zinc, iron, and copper (19). To test whether this observation is reflected in the CaCo-2 Transwell system, we measured their specific contents by inductively coupled plasma MS (ICP-MS). We found the specific contents of manganese, zinc, and iron in CaCo-2 ⌬ZIP14 monolayers to be significantly lower compared with those in CaCo-2 WT cells (Fig. 2D). We ascribe this discrepancy to the in vivo situation to limitations of our Transwell system. In vivo enterocytes can acquire nutrients from both the intestinal lumen and the arterial blood. Transporters functioning in metal uptake from the luminal side of the proximal intestine, such as DMT1, are optimally adjusted to the acidic milieu of the intestinal tract (28 -30). In the Transwell system, however, the pH in the apical chamber is maintained at 7.4 during regular culturing, impairing efficient metal uptake from this side.
To confirm previous results on the localization of ZIP14 in enterocytes (21,31) we studied the localization of ZIP14 in polarized CaCo-2 WT monolayers. Basolateral and apical surface proteins were isolated by a biotinylation approach. Immunoblotting for ZIP14 demonstrated the enrichment of ZIP14 at the basolateral membrane of CaCo-2 WT cells ( Fig. 2E and Fig.   S2). This localization of ZIP14 in enterocytes supports a role of ZIP14 in the uptake of manganese from the blood.

ZIP14 is the major transporter responsible for manganese uptake at the basolateral membrane
To establish the function of ZIP14 in the manganese metabolism of enterocytes, we first studied the consequences of ZIP14 KO on basolateral 54 Mn accumulation. If not stated otherwise, during the experiments the extracellular pH was maintained at 6.1 in the apical and 7.4 in the basolateral compartment to mimic the physiologic situation, i.e. the acidic intraluminal pH of the proximal small intestine and the normal blood pH. Upon addition of 0.1 M 54 Mn, which is close to the low manganese concentrations in mammalian serum (0.01-0.04 M; Ref. 32) to the basolateral chamber, CaCo-2 WT cells rapidly accumulated 54 Mn up to 180 min, after which cellular 54 Mn contents had reached a steady state (Fig. 3A). Basolateral 54 Mn accumulation by CaCo-2 ⌬ZIP14 cells was severely compromised, identifying ZIP14 as the primary transporter mediating manganese uptake at the basolateral membrane of CaCo-2 cells. The residual basolateral 54 Mn accumulation observed in CaCo-2 ⌬ZIP14 cells suggests the presence of an alternative, low-capacity entry route for manganese at the basolateral membrane.
Next we examined the pH dependence of basolateral manganese accumulation by measuring the amount of 54 Mn accumulated by CaCo-2 cells incubated in media at pH 6.1, pH 6.8, or pH 7.4 for 4 h. These experiments were performed by adjusting the incubation media to the same pH in both the basolateral and apical compartments. Basolateral 54 Mn accumulation by CaCo-2 WT cells was maximal at pH 7.4 and decreased with decreasing pH (Fig. 3B). Although CaCo-2 WT cells still accumulated substantial amounts of 54 Mn at pH 6.8, the basolateral 54 Mn accumulation in these cells decreased by ϳ90% at pH 6.1. Consistent with ZIP14 being the major transporter mediating manganese uptake at the basolateral membrane of CaCo-2 WT cells, similar pH dependences have been reported for ZIP14mediated iron and zinc transport (33,34). Opposite to what we observed for CaCo-2 WT cells, basolateral 54 Mn accumulation by CaCo-2 ⌬ZIP14 cells was maximal at pH 6.1 and lower at pH 6.8 and 7.4.

ZIP14 deletion does not substantially affect apical manganese uptake
In contrast to the striking effect of ZIP14 KO observed for basolateral 54 Mn accumulation, 54 Mn accumulation from the apical chamber was observed to be rather similar in both cell lines. Within 6 h both CaCo-2 WT and CaCo-2 ⌬ZIP14 cells accumulated ϳ30 pmol ϫ mg Ϫ1 54 Mn from the apical chamber (Fig.  3C). Although apical 54 Mn accumulation by CaCo-2 ⌬ZIP14 cells proceeded at slightly faster pace between 30 min and 3 h of incubation, 54 Mn contents in CaCo-2 ⌬ZIP14 cells approached a steady state thereafter, whereas 54 Mn contents in CaCo-2 WT cells continued to rise at almost constant speed for up to 6 h. Manganese accumulation by CaCo-2 cells is a complex process involving the uptake and export of manganese at the apical side, its intracellular sequestration and transport, and its basolateral uptake and export. The accelerated accumulation of 54 Mn from ZIP14 knockout enhances manganese absorption the apical side by CaCo-2 ⌬ZIP14 cells compared with CaCo-2 WT cells between 30 min and 3 h may therefore be explained by several mechanisms, including a ZIP14-dependent export process of freshly acquired manganese (either at the basolateral or apical side) that is activated when cellular manganese levels rise above a certain threshold.
Apical 54 Mn accumulation exhibited the same dependence on the extracellular pH in both CaCo-2 WT and CaCo-2 ⌬ZIP14 cells, being maximal at pH 6.1 and decreasing with increasing pH (Fig. 3D). Although this is relevant under physiological conditions, because the pH in the proximal intestine is rather acidic, the mechanisms mediating manganese uptake from the A, confocal microscopy of CaCo-2 WT and CaCo-2 ⌬ZIP14 cells co-immunostained for Na ϩ ,K ϩ -ATPase and ZO-1. The nuclei were stained with 4Ј,6-diamidino-2-phenylindole dihydrochloride. The four upper panels show the orthogonal maximum projections of 42-and 69-frame stacks for CaCo-2 WT and CaCo-2 ⌬ZIP14 cells, respectively. The two bottom panels show a perpendicular projection of merged Na ϩ ,K ϩ -ATPase (red) and ZO-1 (green) proteins with nuclei (blue). The scale bars (50 m) in the upper panels also apply to the bottom panels. B-D, total protein content (B), apparent permeability (P app ) of Lucifer yellow (C), and specific contents of manganese, zinc, iron, and copper of CaCo-2 WT (WT) and CaCo-2 ⌬ZIP14 (⌬ZIP14) Transwell cultures (D). The data are presented as means Ϯ S.D. from 13 (B and C) or 3 (D) independent prepared cultures. Statistical analysis was performed using Student's t test. **, p Ͻ 0.01; ***, p Ͻ 0.001. E, CaCo-2 WT Transwell cultures were subjected to basolateral or apical surface biotinylation with cell membrane-impermeable Sulfo-NHS-SS-biotin. Biotin-labeled cell-surface proteins (Surface) were isolated from whole-cell lysates (Total) using NeutrAvidin-agarose beads. 45 g of whole-cell lysates (Total) and 15 g of biotin-labeled cell-surface proteins (Surface) were analyzed by immunoblotting for ZIP14. Na ϩ ,K ϩ -ATPase served as the basolateral marker and DMT1 as the apical marker.

ZIP14 knockout enhances manganese absorption
intestinal lumen remain unknown. DMT-1, which is strongly expressed at the intestinal brush border (35) and at the apical membrane of CaCo-2 cells (Fig. 2E), had long been regarded as a potential candidate. However, recently Shawki et al. (36) reported that DMT-1 is not required for manganese absorption in intestine-specific Dmt-1 KO mice, contrasting previous observations made in the Belgrade rat model (37, 38).

ZIP14 restricts manganese absorption
Having established that ZIP14 is the major transporter mediating basolateral manganese uptake and that ZIP14 deletion does not substantially affect manganese uptake from the apical membrane of CaCo-2 cells, we next investigated the transport of manganese across CaCo-2 cell monolayers in both basolateral-to-apical and apical-to-basolateral directions. Upon application of 0.1 M 54 Mn to the basolateral side of CaCo-2 WT cells, the amounts of 54 Mn detected in the apical chamber rapidly increased with time (Fig. 4A). Following an exponential phase that lasted ϳ90 min, the amounts of 54 Mn on the apical side increased almost linearly for up to 240 min. In contrast, CaCo-2 ⌬ZIP14 cells transported only minimal amounts of 54 Mn from the basolateral to apical compartment within 240 min. This result could be expected because CaCo-2 ⌬ZIP14 cells lack substantial basolateral manganese uptake activity.
After addition of 0.1 M 54 Mn to the apical chamber to initiate the apical-to-basolateral transport, we observed a slow, exponential rise in the amounts of 54 Mn detected in the apical chamber in the first 150 min followed by a linear increase for up to 240 min for CaCo-2 WT cells (Fig. 4B). The time dependence of apical-to-basolateral manganese transport by CaCo-2 ⌬ZIP14 cells displayed a similar pattern but with significantly higher amounts of 54 Mn accumulated in the basolateral compartment. The increased apical-to-basolateral manganese transport by CaCo-2 ⌬ZIP14 cells compared with CaCo2 WT cells strongly suggests that intestinal ZIP14 is required to restrict manganese absorption.
ZIP14 has been described as a broad-scope metal transporter, capable of mediating the cellular uptake of several metals including iron (33,34). Hence, we tested the consequences of ZIP14 deletion on iron accumulation and transport in CaCo-2 cell monolayers upon addition of 1 M 59 Fe 2ϩ or 59 Fe 3ϩ . Consistent with the function of ZIP14 as a Fe 2ϩ transporter, basolateral 59 Fe 2ϩ accumulation was lowered by 30% in CaCo-2 ⌬ZIP14 cells, whereas 59 Fe 3ϩ accumulation did not differ significantly from CaCo-2 WT cells (Fig. S3A). In contrast to manganese transport, the lack of ZIP14 did not affect the apical-to-basolateral transport of 59 Fe 2ϩ or 59 Fe 3ϩ (Fig. S3, C and D), most likely because iron is released from the cells into the media in its ferric state. This result suggests that in the CaCo-2 Transwell system, functional ZIP14 specifically restricts manganese absorption.

ZIP14 mediates manganese reuptake from the basolateral membrane to restrict manganese absorption
In CaCo-2 cells, ZIP14 restricts the flux of manganese in the absorptive direction (Fig. 4B), but how is it accomplished?

ZIP14 knockout enhances manganese absorption
ZIP14 is localized to the basolateral membrane of CaCo-2 cells (Fig. 2E) and is required for efficient manganese uptake from the basolateral side (Fig. 3, A and B). Accordingly, direct reuptake of freshly absorbed manganese is a plausible mechanism by which ZIP14 may restrict manganese absorption. If so, inhibiting basolateral manganese uptake should increase apical-to-basolateral manganese transport in CaCo-2 WT cells, but not in CaCo-2 ⌬ZIP14 cells. We used three approaches to minimize basolateral manganese reuptake.
First, because ZIP14-mediated 54 Mn uptake is highest at pH 7.4 and decreases by 95% at pH 6.1 (Fig. 3B), we studied the transport of 0.1 M 54 Mn from the apical to basolateral compartment with the pH in the basolateral chamber being maintained at pH 7.4 as the normal condition or adjusted to pH 6.1. For CaCo-2 WT cells, 54 Mn accumulation in the basolateral chamber was found to be approximately five times higher at a basolateral pH of 6.1 compared with a basolateral pH of 7.4 (Fig.  5A). In contrast, apical-to-basolateral manganese transport by CaCo-2 ⌬ZIP14 cells slightly decreased with lowering the basolateral pH from pH 7.4 to pH 6.1 (Fig. 5B). These observations are consistent with the pH dependence of basolateral manganese accumulation (Fig. 3B) and suggest that manganese transport in the apical-to-basolateral direction is limited by basolateral manganese uptake.
Apical-to-basolateral manganese transport requires ϳ150 min to become fully established (Fig. 4B). To exclude that the modulation of basolateral mediated manganese uptake by altering the basolateral pH induces changes in the recruitment of transporters required for the export of intracellular manganese at the basolateral side, we removed 54 Mn from the apical compartment after 4 h of 54 Mn preloading and followed 54 Mn accumulation in the basolateral compartment for another 3 h, with the pH of the basolateral compartment either maintained at the same pH as in the first 4 h or switched to pH 6.1 or pH 7.4 ( Fig.  5, C-F). Alteration of the basolateral pH caused a rapid response in the apical-to-basolateral transport rates. The immediate effect of the basolateral pH switch on the apical-tobasolateral manganese transport in CaCo-2 cells suggests that the elevated manganese absorption at lower pH observed in CaCo-2 WT cells is indeed due to impaired ZIP14 mediated manganese uptake at low pH rather than due to alterations in transporter abundance at the basolateral membrane.
Second, we used excess zinc to compete with 54 Mn reuptake from the basolateral chamber. Zinc has previously been reported as a potent competitive inhibitor of ZIP14-mediated manganese uptake (33). However, initial experiments revealed that a 10-fold excess of zinc does not inhibit basolateral 54 Mn accumulation by CaCo-2 WT cells when Dulbecco's modified Eagle's medium (DMEM) was used as the experimental medium (Fig. S4A). In contrast, when salt-based incubation buffer was used instead of DMEM, excess zinc strongly impaired basolateral 54 Mn accumulation in CaCo-2 WT cells (Fig. S4B). Thus, to study the effect of zinc on apical-to-basolateral transport, CaCo-2 monolayers were first loaded for 4 h with 0.1 M 54 Mn from the apical side, after which the accumulation of 54 Mn in the basolateral compartment was followed for 90 min using salt-based incubation buffer with or without 2 M zinc. The presence of 2 M zinc strongly enhanced 54 Mn accumulation in the basolateral compartment of CaCo-2 WT Transwell cultures (Fig. 6A), suggesting that zinc impairs the direct reuptake of freshly absorbed manganese by competing with 54 Mn for uptake via ZIP14. Consistent with this view, 54 Mn accumulation in the basolateral compartment of CaCo-2 ⌬ZIP14 Transwell cultures (Fig. 6B) was not altered by zinc.
Third, we sought to prevent ZIP14-mediated manganese reuptake by chelating freshly absorbed manganese with EDTA or desferrioxamine (DFO), which are strong chelators for Mn(II) and Mn(III), respectively (39,40). Because both EDTA and DFO are membrane-impermeable (41), the intracellular manganese trafficking would not be disturbed. To carry out the experiment, CaCo-2 monolayers were first loaded for 4 h with 0.1 M 54 Mn from the apical side, after which the accumulation of 54 Mn in the basolateral compartment was followed for 90 min using salt-based incubation buffer containing either no chelator, 100 M CaNa 2 EDTA, or 1 mM DFO. The presence of both EDTA or DFO caused an increased accumulation of 54 Mn

ZIP14 knockout enhances manganese absorption
in the basolateral compartment of CaCo-2 WT Transwell cultures (Fig. 6, C and E) but not in CaCo-2 ⌬ZIP14 Transwell cultures (Fig. 6, D and F). Because manganese is released close to the plasma membrane, the moderate effect of both chelators on the accumulation of 54 Mn in the basolateral compartment of CaCo-2 WT Transwell cultures is most likely due to limited access of EDTA and DFO to manganese prior to its reuptake. Nevertheless, these results further support our proposed mechanism that ZIP14 restricts manganese absorption by direct reuptake of freshly absorbed manganese.

Intestine-specific deletion of Zip14 in mice leads to manganese overload
To test the physiologic relevance of our findings, we generated mice carrying Zip14 conditional alleles (Zip14 flox/flox mice) by flanking exons 3 and 4 of the Zip14 with loxP sites (Fig.  7A). The correct insertion of loxP sites was verified by genotyping (Fig. 7B). We then generated intestine-specific Zip14 KO mice by crossing Zip14 flox/flox mice with Vil-Cre mice. For comparison, we also generated liver-specific Zip14 KO mice. Tissue-specific inactivation of Zip14 was confirmed by immunoblot analysis (Fig. 7C and Fig. S5).
We used ICP-MS to determine manganese levels in whole blood, livers, and brains from Zip14 flox/flox mice, intestinespecific, and liver-specific Zip14 KO mice at 3 weeks of age.
We observed a clear difference in the hepatic manganese levels from liver-specific Zip14 KO mice and intestine-specific Zip14 KO mice (Fig. 7D): consistent with previous results (19), hepatic manganese contents were significantly reduced in liver-specific Zip14 KO mice; in contrast, manganese levels in livers from intestine-specific Zip14 KO mice were more than two times that of controls, revealing that loss of ZIP14 in the intestine increases body manganese burden. This observation establishes the importance of the intestinal ZIP14 in the control of systemic manganese homeostasis and is in full agreement with our discovery that deletion of ZIP14 in CaCo-2 cells results in increased apical-to-basolateral manganese transport.
Although manganese concentrations in the blood of intestine-specific Zip14 KO mice did not increase (Fig. 7D), brain manganese contents were double that of control and liver-specific Zip14 KO mice at 3 weeks old. An explanation could be that manganese homeostasis can still be regulated through hepatobiliary excretion in intestine-specific Zip14 KO mice, but the liver cannot clear all of the excess manganese entering the portal blood, and elevated manganese entering the systemic circulation is readily absorbed by other tissues, including the brain. Further studies are required to evaluate the manganese contents in other tissues of intestine-specific Zip14 KO mice

ZIP14 knockout enhances manganese absorption
and to examine the consequences of Zip14 deletion in both the liver and intestine.

Discussion
In the present study, we identified ZIP14 as the major transporter mediating basolateral manganese uptake in enterocytes. Moreover, we provide in vitro and in vivo evidence that this function of ZIP14 is required for the control of manganese homeostasis.
CaCo-2 cell cultures provide a well-established model system to study the uptake and the absorptive and secretory fluxes of drugs and nutrients in the intestinal epithelium (25). Genetically modified CaCo-2 cell lines have previously been used to study the uptake and transport of metals. Just recently, a stable CaCo-2 cell line overexpressing ZnT10 has been employed to prove the ability of this transporter to mediate the efflux of cellular manganese at the apical membrane (52). RNAi-mediated gene knockdown has been used to study the functions of ZnT1, ferroportin, and DMT1 in CaCo-2 cell monolayers (42,43). A drawback of RNAi-mediated gene knockdown is the possibility of residual gene expression that may mask potential phenotypes, a complication that we avoided by generating a ZIP14 KO CaCo-2 cell line using the CRISPR/Cas9 system.
In contrast to previous studies investigating manganese uptake and manganese transport by CaCo-2 cells (44 -46), we maintained a pH of 6.1 in the apical and a pH of 7.4 in the basolateral compartment to mimic the physiologic situation. Such pH gradient is routinely used in studies investigating iron uptake and iron transport in CaCo-2 cells and is of the uttermost importance when studying apical manganese uptake and apical-to-basolateral manganese transport because substantial manganese transport in the absorptive direction occurs only at acidic apical pH (Fig. S6). To account for the low manganese Systemic manganese homeostasis is maintained by the regulation of both intestinal manganese absorption and hepatobiliary manganese excretion (3)(4)(5)(6). The accumulation of excess manganese in blood and several organs in individuals with homozygous mutations in ZIP14 (10, 16 -18) and global Zip14 KO mice (19 -22) indicates a crucial requirement of ZIP14 in systemic manganese homeostasis. ZIP14 has convincingly been shown to mediate hepatobiliary manganese excretion in concert with ZnT10 (20). However, in contrast to the global Zip14 KO mice, liver-specific Zip14 KO mice do not hyperaccumu-

ZIP14 knockout enhances manganese absorption
late manganese (19), indicating that ZIP14 is involved in extrahepatic processes required for the maintenance of systemic manganese homeostasis.
In the present study, we identified ZIP14 as the primary transporter mediating efficient manganese uptake from the basolateral side of enterocytes. This function of ZIP14 is consistent with its localization in the intestine and CaCo-2 monolayers (21,31) and the reported pH dependences of ZIP14-mediated iron and zinc uptake (33,34). Our results also indicate that CaCo-2 cells possess alternative systems to acquire manganese from the basolateral compartment that become evident in CaCo2 ⌬ZIP14 cultures; however, because of the low capacity, these mechanisms are unlikely to compensate for the loss of ZIP14 function in vivo.
Apical manganese uptake does not differ substantially in CaCo-2 WT and CaCo2 ⌬ZIP14 cultures, but ZIP14 inactivation strongly enhances the flux of manganese in the absorptive direction, suggesting that functional ZIP14 is required to limit manganese absorption. Mechanistically, we demonstrate that inhibiting ZIP14-mediated basolateral manganese uptake by lowering the basolateral pH, zinc competition, and chelation of freshly absorbed manganese all increase manganese transport in the absorptive direction. These observations strongly support that intestinal ZIP14 contributes to the maintenance of

ZIP14 knockout enhances manganese absorption
systemic manganese homeostasis by restricting dietary manganese absorption via direct reuptake of freshly absorbed manganese. Such function of intestinal ZIP14 would not only explain the manganese hyperaccumulation observed in individuals lacking functional ZIP14 but also the lack of manganese accumulation in the liver-specific Zip14 KO mice.
Providing evidence that ZIP14 restricts dietary manganese absorption in vivo, intestine-specific Zip14 KO mice display elevated manganese contents in the liver and brain, indicating that enterocyte-specific KO of Zip14 causes manganese overload. However, these mice do not develop increased blood manganese concentrations at 3 weeks of age because the liver control system is still in place and excess manganese is likely to be cleared from the blood by extrahepatic tissues, including the brain.
In summary, we show that ZIP14 is crucial for efficient manganese uptake from the basolateral membrane of enterocytes. We provide evidence that one potential mechanism by which ZIP14 restricts manganese absorption and limits whole body manganese burden is through direct reuptake of freshly absorbed manganese. Moreover, our in vivo data provide strong evidence that intestinal ZIP14 plays an important role in maintaining systemic manganese homeostasis.

Cell cultures
CaCo-2 (ATCC HTB-37 TM ) cells were maintained in growth medium (80% DMEM containing 1 mM pyruvate; Corning, Corning, NY) supplemented with 3.7 g ϫ liter Ϫ1 NaHCO 3 , 1ϫ nonessential amino acids (Thermo Fisher Scientific), 100 units ϫ ml Ϫ1 penicillin, 100 g ϫ ml Ϫ1 streptomycin (Thermo Fisher Scientific), and 20% fetal bovine serum (VWR, Radnor, PA) at 37°C and 5% CO 2 in the humidified atmosphere of an incubator. The manganese concentration of the growth medium was determined to be 0.056 Ϯ 0.005 M by ICP-MS as described under "ICP-MS analysis of metals" (Fig. S7). The cells were split every 3-4 days following detachment by treatment with 0.25% trypsin in PBS without calcium and magnesium and containing 0.2% EDTA (Thermo Fisher Scientific). For routine cell culture, the cells were seeded in cell culture dishes of 55 cm 2 (Thermo Fisher Scientific) at a density of 0.1 ϫ 10 4 (CaCo-2 WT ) or 0.14 ϫ 10 4 (CaCo2 ⌬ZIP14 ) cells per cm 2 . For preparation of CaCo-2 cell monolayers, 1 ϫ 10 5 cells/cm 2 were seeded in Transwell-Clear TM inserts with 0.4-m pore size of 4.67 or 1.12 cm 2 surface area (both from Corning) that had been incubated with growth medium for 24 h prior to seeding. Inserts of 4.67 (1.12) cm 2 were placed in 6-well (12-well) plates with 2.6 (1.5) ml of growth medium in the well and 1.5 (0.5) ml in the insert. The growth medium was changed every 2-3 days. CaCo-2 cell monolayers were used for experiments between 21 and 27 days after seeding.

Generation of a CaCo-2 ZIP14 KO cell line
The CaCo-2 ZIP14 KO cell line was generated using the CRISPR-Cas9 system (49). The guide RNA sequence (5Ј-GTG CAG CAG CAG CAG CTT CA-3Ј), complementary to a sequence in proximity to the start codon of the ZIP14 gene, was cloned into the pX330-U6-Chimeric_BB-CBh-hSpCas9 vector (50). CaCo-2 cells were transfected with ZIP14-targeting vector using Lipofectamine TM 3000 (Thermo Fisher Scientific). Briefly, on the day of the transfection, CaCo-2 cells were seeded in 12-well plates (Thermo Fisher Scientific) at a density of 0.2 ϫ 10 4 cell/cm 2 and incubated for 6 h in growth medium without penicillin and streptomycin. The medium was changed to growth medium, and the DNA-lipid complex (1 g of plasmid DNA, 2 l of P3000 TM reagent, 3 l of Lipofectamine TM 3000 reagent, 95 l of DMEM), prepared 3 h in advance according to the protocol provided by the manufacturer, was added dropwise to the cells. The cells were grown for 72 h, after which the medium was changed to growth medium containing 7.5 g ϫ ml Ϫ1 puromycine for selection. After 48 h the cells contained in one well were split into four culture dishes of 55 cm 2 and cultured for 3 weeks in growth medium. Single-cell clones were transferred to 12-well plates for screening by immunoblot analysis with a human ZIP14-specific antibody described under "Antibodies." Cell clones lacking ZIP14 expression were subcloned. ZIP14 inactivation in selected subclones was confirmed by immunoblot analysis and DNA sequencing.

Animals
All mice were maintained on a NIH-31 irradiated traditional rodent diet (Teklad 7913; Envigo, Indianapolis, IN) and housed in the laboratory animal facility at the University of Arizona. Procedures for animal experiments were approved by the institutional animal care and use committee. Mice carrying Zip14 conditional alleles (Zip14 flox/flox mice) were generated by flanking exons 3 and 4 of Zip14 gene with loxP sites (Cyagen, Santa Clara, CA). The transgenic mice expressing Cre recombinase under the intestine-specific villin promotor (Vil-Cre mice) and liver-specific albumin promoter (Alb-Cre mice) were purchased from the Jackson Laboratory. All mice are on the C57BL/6 background. Genotyping procedures were performed by using mouse tail snipping and the Mouse Direct PCR kit (Bimake, Houston, TX). The mice were sacrificed after anesthetizing with ketamine/xylazine at 3 weeks of age. Tissues were collected and immediately frozen in liquid nitrogen and stored at Ϫ80°C until further analyses.

Antibodies
Because antibodies against mouse ZIP14 (mZIP14) and human ZIP14 (hZIP14) are not commercially available, we employed the approach of expressing glutathione S-transferase (GST) fusion proteins in Escherichia coli to produce and purify the immunogens for the production of antibodies. To construct a vector carrying mZIP14 or hZIP14 fusion protein, the sequence encoding N-terminal serine 31 to proline 146 of mZIP14 or serine 32 to proline 148 of hZIP14 was PCR-amplified and cloned into pGEX-3X (Addgene, Watertown, MA) vector using BamHI-EcoRI linkers. The fusion protein was isolated from E. coli and purified by affinity chromatography on gluthathione-Sepharose 4B (GE Healthcare). The immunization procedures were performed by the Pocono Rabbit Farm & Laboratory (Canadensis, PA). The antisera obtained from the test bleeds were analyzed for ZIP14 recognition by immunoblotting. Antisera were cleared by GSH-Sepharose crosslinked with GST to remove anti-GST antibody. The cleared ZIP14 knockout enhances manganese absorption flow-through fractions were collected and used to purify anti-hZIP14 or anti-mZIP14 antibody using GSH-Sepharose crosslinked with GST-mZIP14 or GST-hZIP14 fusion protein (Fig. S8).

Immunoblot analysis
Cells and mouse tissues were lysed in NETT buffer (150 mM NaCl, 5 mM EDTA, 10 mM Tris, 1% Triton X-100, and 1ϫ protease inhibitor mixture (Bimake, Houston, TX), pH 7.4). The nuclei were removed by centrifugation (15,000 ϫ g at 4°C for 15 min). Protein concentrations of cell lysates were determined using the RC DC TM protein assay (Bio-Rad). Cell lysates were mixed with 1ϫ Laemmli buffer and incubated for 30 min at 37°C. The proteins were separated electrophoretically on an SDS/10% polyacrylamide gel and transferred to nitrocellulose membranes (GVS, Sanford, ME). Following blocking of unspecific binding sites with 5% (w/v) nonfat dry milk in TBS with Tween 20 (TBS-T; 10 mM Tris/HCl, 150 mM NaCl, 0.1% (v/v), 1 ml of Tween 20, pH 7.5), membranes were probed at 4°C overnight with rabbit anti-hZIP14 (1:3000), anti-mZIP14, or anti-DMT1 (1:5000) antibodies in blocking buffer. The membranes were washed four times with TBS-T (5 min each) and subsequently incubated for 1 h at room temperature with HRP-conjugated goat anti-rabbit secondary antibodies (1:5000). After two washes with TBS-T and TBS (5 min each), the blots were developed using enhanced chemiluminescence (SuperSignal West Pico, Thermo Fisher Scientific) and the ChemiDoc TM MP Imaging System (Bio-Rad). To confirm equivalent loading, blots were stripped for 15 min in Restore PLUS Western blotting stripping buffer (Thermo Fisher Scientific), blocked for 1 h in blocking buffer, and reprobed with HRP-conjugated antibodies directed against GAPDH (1:20,000) and/or ␤-ACTIN (1:20,000). Mouse anti-Na ϩ ,K ϩ -ATPase (1:3000) followed by HRP-conjugated secondary antibodies (1:5000) served as loading control for plasma membrane proteins. The blots were visualized and quantified using the ChemiDoc MP imaging system with Image Lab software (Bio-Rad).

Isolation of genomic DNA, PCR, and sequencing
Genomic DNA was isolated using the Quick-DNA TM Miniprep Plus Kit (Zymo Research, Irvine, CA) according to the protocol provided by the manufacturer. The guide RNA targeting region in exon 2 of the ZIP14 gene was amplified with DNAspecific primers (forward primer, 5Ј-GAG CAG AGA AGC AGA GAC TGA-3Ј; reverse primer, 5Ј-ACG GTA AGG CTC CCC TGT-3Ј). PCR products were isolated by agarose gel electrophoresis and purified with the Wizard SV Gel and PCR Clean-Up System (Promega Corporation, Madison, WI). The purified PCR products were sent for sequencing to MCLAB (San Francisco, CA).

ICP-MS analysis of metals
CaCo-2 cell monolayers grown on Transwell inserts were washed twice with PBS, after which membranes were excised with a scalpel, transferred to 1.5-ml reaction tubes, and frozen in liquid nitrogen. The cells were lysed by repeated freeze-thaw in 1 ml of MilliQ-water, and an aliquot of 25 l was used to determine the cellular protein content according to the Lowry method (51) using BSA as a standard. An aliquot of 850 l of the cell lysate was incubated with 300 l of HNO 3 (67-70%; BDH ARISTAR PLUS, VWR) at 85°C overnight followed by 2 h at 95°C and diluted to 7 ml with MilliQ-water. For the determination of the metal content in media, 10 ml of media were digested with 450 l of HNO 3 . The digested samples and mice tissues were sent for ICP-MS analysis to the Arizona Laboratory for Emerging Contaminants (Tucson, AZ).

Immunocytochemistry and confocal microscopy
CaCo-2 cell monolayers grown on Transwell inserts were washed twice with 2.6 ml (basolateral compartment) and 1.5 ml (apical compartment) of ice-cold PBS Ca/Mg (PBS supplemented with 0.5 mM MgCl 2 and 0.9 mM CaCl 2 , pH 7.2) and subsequently fixed with methanol for 15 min at Ϫ20°C. After adding acetone (prechilled to Ϫ20°C) for 30 s, the cells were washed thrice (5 min each) with PBS Ca/Mg . The membranes were excised with a scalpel, transferred to fresh 6-well plates, and blocked with 1% BSA in PBS Ca/Mg for 2 h at room temperature. If not stated otherwise, the cells were washed thrice with PBS Ca/Mg in 5-min intervals between the different steps of the following staining procedure. Incubation of the membranes with rabbit anti-ZO-1 (1:100) and mouse anti-Na ϩ ,K ϩ -ATPase (1:100) was carried out at 4°C overnight in humidified atmosphere followed by an incubation with Alexa Fluor 488conjugated goat anti-rabbit (1:500) and Alexa Fluor 594conjugated goat anti-mouse antibodies (1:500) in 1% BSA in PBS Ca/Mg for 1 h at room temperature. To visualize the nuclei, the cells were exposed to DAPI (1 g ϫ ml Ϫ1 in PBS) for 5 min at room temperature. The membranes were mounted on glass slides with cells facing up in ProLong Diamond antifade mountant (Thermo Fisher Scientific). A coverslip was placed on top of the cells and sealed with nail polish. The images were captured by a Zeiss LSM880 inverted confocal microscope with a 63ϫ oil objective at the Marley Imaging Core of the University of Arizona (Tucson, AZ).

Integrity of the monolayer
Monolayer integrity of CaCo-2 cells was determined by the LY rejection assay. The cells were washed twice with 2.6 ml (basolateral compartment) and 1.5 ml (apical compartment) of PBS Ca/Mg , after which the inserts were transferred to fresh 6-well plates. 2.6 ml of prewarmed (37°C) salt-based incubation buffer (145 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 0.8 mM Na 2 HPO 4 , 50 mM glucose, 20 mM PIPES) ZIP14 knockout enhances manganese absorption adjusted to pH 7.4 and 1.5 ml of prewarmed (37°C) incubation buffer containing 12.5 mg ϫ ml Ϫ1 LY (Sigma-Aldrich) adjusted to pH 6.1 were added to the basolateral and apical compartment, respectively, and the cells were incubated for 4 h at 37°C in an incubator. An aliquot of 1 ml of incubation buffer was removed from the basolateral compartment every 60 min and replaced by the same volume of fresh incubation buffer adjusted to pH 7.4. At the end of the experiment, the media from both compartments were collected. Standards containing 12.5 g ϫ ml Ϫ1 to 6.1 ng ϫ ml Ϫ1 were prepared in incubation buffer adjusted to pH 7.4. Aliquots of 150 l of standard or sample were transferred into a well of a black microtiter plate, and fluorescence was quantified using a BioTek Synergy 2 plate reader (BioTek Instruments Inc., Winooski, VT). The pH of the media at the start and the end of the experiment was measured with a Sartorius pHCore pH meter (Sartorius, Lab Instruments GmbH & Co. KG, Goettingen, Germany).

Cell-surface biotinylation
Cell-surface biotinylation was performed on CaCo-2 cell monolayers grown on Transwell inserts. The cells were washed twice with 2.6 ml (basolateral compartment) and 1.5 ml (apical compartment) PBS Ca/Mg prewarmed to 37°C and twice with PBS Ca/Mg prechilled to 4°C, after which the insert was transferred to a fresh 6-well plate. For biotinylation of basolateral (apical) surface proteins, 1.5 ml (1 ml) of freshly prepared, icecold solution containing 1 mg ϫ ml Ϫ1 Sulfo-NHS-SS-biotin (Thermo Fisher Scientific) in PBS Ca/Mg was added to the basolateral (apical) compartment, and 1 ml (1.5 ml) of ice-cold PBS Ca/Mg was added to the apical (basolateral) compartment. The cells were incubated on ice in an incubator at 4°C for 30 min with gentle rocking. The cells were then washed twice with 2.6 ml (basolateral compartment) and 1.5 ml (apical compartment) PBS Ca/Mg prechilled to 4°C. The reaction was quenched by incubating the cells with 2.6 ml (basolateral compartment) and 1.5 ml (apical compartment) of 100 mM glycine in PBS Ca/Mg for 20 min on ice in an incubator at 4°C with gentle rocking. Again the cells were washed twice with 2.6 ml (basolateral compartment) and 1.5 ml (apical compartment) of PBS Ca/Mg prechilled to 4°C. The insert was transferred to a fresh 6-well plate, and the membrane was excised with a scalpel. The cells were lysed in 1 ml of NETT buffer, and the lysate was transferred into 1.5-ml reaction tubes. The lysate was chilled on ice for 20 min with vortexing every 5 min, sonicated with 10 strikes (2 s, amplitude 20) of a Q55 sonicator (Qsonica, Newton, CT) and centrifuged (15,000 ϫ g, 4°C, 15 min). The supernatant was transferred into fresh 1.5-ml reaction tubes. An aliquot of 150 l (total protein) was frozen in liquid nitrogen and stored at Ϫ80°C for later immunoblot analysis, and 850 l of the lysates were transferred to Pierce TM centrifuge columns that had been preloaded with 100 l of Pierce TM high capacity NeutrAvidin TM agarose (both from Thermo Fisher Scientific) according to the manufacturer's protocol. The samples were incubated overnight at 4°C with slight rocking motion, after which the supernatant was removed by centrifugation for 1 min at room temperature and 1500 ϫ g. The agarose beads were washed once with NETT containing protease inhibitor, two times with NETT buffer without protease inhibitor, two times with salt wash buffer (350 mM NaCl, 5 mM EDTA, 0.1% Triton X-100 in PBS, pH 7.4), and three times with NET buffer (150 mM NaCl, 5 mM EDTA, 10 mM Tris, pH 7.4). Between the individual washing steps, the supernatant was removed by centrifugation for 1 min at room temperature and 1500 ϫ g. To elute the biotinylated proteins, 100 l of 1ϫ sample buffer (1.7% (w/v) SDS, 5% (v/v) glycerol, 150 mM DTT, 58 mM Tris, pH 6.8) was added to the agarose beads, and the samples were incubated at 4°C overnight. On the next day, the centrifuge columns were placed in fresh 1.5-ml reaction tubes, incubated for 30 min at 37°C, and centrifuged for 2 min at room temperature and 1500 ϫ g. The eluate (surface protein) was transferred to fresh 1.5-ml reaction tubes, frozen in liquid nitrogen, and stored at Ϫ80°C for later immunoblot analysis.

Preparation of radiolabeled metal solutions
Radiolabeled metal solutions were prepared on the day of the experiment. 5-Fold concentrated (0.5 M) radiolabeled manganese solutions were prepared from 54 MnCl 2 (PerkinElmer Life Sciences) complexed to citrate prior to addition to DMEM. 5-Fold concentrated (5 M) radiolabeled Fe(II) solutions were prepared from 59 FeCl 3 (PerkinElmer Life Sciences) that had been reduced by 5 mM ascorbate prior to addition to DMEM containing 5 mM ascorbate. 5-Fold concentrated (5 M) radiolabeled Fe(III) solutions were prepared from 59 FeCl 3 that had been complexed to citrate prior to addition to DMEM.

Metal uptake and transport
Metal uptake and transport was studied in CaCo-2 cell monolayers grown on Transwell inserts. During the experiments, the cells were incubated at 37°C in the humidified atmosphere of an incubator, and all solutions required before the termination of the experiment were prewarmed to 37°C. At the start of the experiments, the cells were washed twice with 2.6 ml (basolateral compartment) and 1.5 ml (apical compartment) of PBS Ca/Mg . The inserts were then transferred to fresh 6-well plates. If not stated otherwise, 2.6 ml of transport medium (DMEM containing 1 mM pyruvate supplemented with 20 mM PIPES) adjusted to pH 7.4 and 1.5 ml of transport medium adjusted to pH 6.1 were added to the basolateral and apical compartment, respectively, and the cells were preincubated for 30 min. Basolateral (apical) metal uptake and basolateral-to-apical (apical-to-basolateral) transport experiments were initiated by replacing 520 l (300 l) of transport medium of the basolateral (apical) compartment by the same volume of a freshly prepared 5-fold concentrated radiolabeled metal solution, and the cells were incubated for the desired times. When basolateral-to-apical (apical-to-basolateral) transport was studied an aliquot of 1 ml of medium was removed from the apical (basolateral) compartment for ␥-counting by a 2480 WIZARD 2 Automatic Gamma Counter (PerkinElmer Life Sciences) and replaced by the same volume of the media initially added to the compartment at the time points indicated in the figures. The experiments were stopped by aspirating the media from both compartments and washing the cells three times with ice-cold PBS supplemented with 1 mM EDTA. The cells were lysed in 1 ml of 0.5 M NaOH, and an aliquot of 25 l was used to determine the cellular protein content according to the

ZIP14 knockout enhances manganese absorption
Lowry method (51) using BSA as a standard. An aliquot of 200 -800 l of the lysate were used to quantify the cellular contents of 54 Mn and 59 Fe by ␥-counting.

Statistical analysis
Significance of differences between two sets of data were analyzed using Student's t test. Comparisons between multiple sets of data were performed using one-or two-way analysis of variance (ANOVA) followed by the Bonferroni post hoc test: *, p Ͻ 0.05; **, p Ͻ 0.01; and ***, p Ͻ 0.001. p Ͼ 0.05 was considered as not significant. The PRISM 5 software (GraphPad, La Jolla, CA) was used for statistical analysis.