Paracellin-1, a Renal Tight Junction Protein Required for Paracellular Mg2+ Resorption

See allHide authors and affiliations

Science  02 Jul 1999:
Vol. 285, Issue 5424, pp. 103-106
DOI: 10.1126/science.285.5424.103


Epithelia permit selective and regulated flux from apical to basolateral surfaces by transcellular passage through cells or paracellular flux between cells. Tight junctions constitute the barrier to paracellular conductance; however, little is known about the specific molecules that mediate paracellular permeabilities. Renal magnesium ion (Mg2+) resorption occurs predominantly through a paracellular conductance in the thick ascending limb of Henle (TAL). Here, positional cloning has identified a human gene,paracellin-1 (PCLN-1), mutations in which cause renal Mg2+ wasting. PCLN-1 is located in tight junctions of the TAL and is related to the claudin family of tight junction proteins. These findings provide insight into Mg2+homeostasis, demonstrate the role of a tight junction protein in human disease, and identify an essential component of a selective paracellular conductance.

The composition of fluids on opposite sides of epithelial cell layers is often different, and is maintained by barriers to passage of fluid, electrolytes, nutrients, pathogens, and cells. Cellular membranes provide a transcellular barrier, whereas the tight junctions (zonula occludens) have long been recognized to constitute the intercellular permeability barrier (1). These transcellular and paracellular barriers are not absolute, and they often display selective and regulated conductances (2). Although many mediators of transcellular flux have been characterized, specific determinants of paracellular permeabilities are unknown. Several components of the tight junction have been identified at the molecular level (3), but their specific roles have yet to be determined.

Magnesium is a critical cofactor in a wide variety of biological activities. The renal resorption of Mg2+ occurs predominantly by paracellular flux (4) in the thick ascending limb of Henle (TAL) (5), a process driven by a favorable electrochemical gradient (Fig. 1A). The conductance of this paracellular pathway is highly regulated (6), with renal Mg2+ excretion varying from 0.5 to 80% of the filtered load with low or high serum Mg2+ concentrations, respectively (7). Interestingly, although this paracellular pathway shows relatively high magnesium conductance, it is highly impermeable to water (8).

Figure 1

(A) Paracellular resorption of Mg2+ and Ca2+ in the renal TAL. A cell of the TAL is depicted along with portions of adjacent cells. A lumen-positive potential is established by the Na+-K+-2Cl cotransporter bringing these ions into the cell and the K+ channel ROMK recycling K+ back into the lumen. This positive potential and the favorable chemical gradient provide the driving force for paracellular resorption of Mg2+ and Ca2+. (B) Family structure of kindred K113. Family members with recessive renal hypomagnesemia are indicated by solid symbols and numbered as in (C). (C) Mapping the recessive renal hypomagnesemia locus. Affected persons from kindreds K113 and K114 are shown with genotypes from chromosome 3q marker loci. The distances between adjacent loci are indicated. Homozygous segments are indicated by gray boxes. All of the affected persons show segments of homozygosity within this interval; however, only one marker (D3S1314) is homozygous in all of these individuals, suggesting a location of the disease gene between loci 539-5 and D3S1288. (D) Overlapping BAC clones spanning the interval containing the disease gene (13). The locations of STSs and polymorphic loci are indicated at the top. The location of the disease locus inferred from genetic studies is indicated.

With these observations in mind, a rare autosomal recessive disease, renal hypomagnesemia with hypercalciuria and nephrocalcinosis, is an intriguing entity (9). Affected persons have profound renal Mg2+ wasting, which results in severe hypomagnesemia that is not corrected by oral or intravenous Mg2+supplementation. They also have renal Ca2+ wasting, resulting in renal parenchymal calcification and renal failure. Other features include urinary tract infections, kidney stones, hyperuricemia, and ocular findings. These patients do not display salt wasting, and renal transplantation cures the defects in electrolyte homeostasis. Because the massive renal Mg2+ loss suggests a defect in paracellular resorption, identification of the disease gene could provide insight into the nature of paracellular permeabilities and regulation of magnesium homeostasis.

Twelve kindreds with typical recessive renal hypomagnesemia were recruited for study (10, 11); in 10 families affected persons were the offspring of consanguineous unions. To map the gene (or genes) causing hypomagnesemia in these kindreds, we genotyped polymorphic loci across the human genome and analyzed linkage to the trait (12). Initial analysis of three consanguineous kindreds demonstrated linkage to a segment of chromosome 3q, with affected persons homozygous for many consecutive markers in this interval and a lod score (logarithm of odds ratio for linkage) of 6.8 (11). All other families also supported linkage to this segment, and two provided the basis for a refined location of the disease gene. In kindred K113, 11 distantly related affected members were studied (Fig. 1B). All showed homozygosity on 3q (estimated lod score, >20); however, all shared homozygosity for identical alleles of only two consecutive markers, D3S1314 and 539-5(Fig. 1C). Analysis of kindred K114, with two affected siblings from a second-cousin marriage, further refined the location of the trait locus, with one affected sibling losing homozygosity at locus539-5 (Fig. 1C). Together these findings localize the trait locus to the approximately 1-centimorgan interval flanked by loci539-5 and D3S1288.

We next constructed a physical map of the interval by selection of bacterial artificial chromosome (BAC) clones hybridizing to sequence-tagged sites (STSs) and polymorphic markers from the critical interval (13). The results demonstrate a continuous map spanning about 1 Mb in genomic DNA (Fig. 1D). The flanking markers539-5 and D3S1288 are contained in the physical map and are separated by no more than 500 kb, which shows that the trait locus lies within this short chromosomal interval.

Examination of public databases revealed no known genes in the critical interval. Candidate genes were sought by exon trapping using BACs from the physical map (14). One of the resulting clones, trapped from BAC 1314-14, yielded a product containing a putative 54–amino acid open reading frame (ORF), which ultimately proved to be exon 3 of a gene. The complete ORF of this gene, which we have termed paracellin-1 (PCLN-1), was determined (14) and sequence analysis showed that it encodes a protein of 305 amino acids (Fig. 2A) with four transmembrane domains and intracellular NH2- and COOH-termini (11, 15). The PCLN-1 protein shows sequence and structural similarity to members of the claudin family (Fig. 2A) (15, 16) and is the most distantly related member of this family (Fig. 2B). More than a dozen members of this family have been identified; all localize to tight junctions and appear to bridge the intercellular space by homo- or heterotypic interactions (17). PCLN-1 shows 10 to 18% amino acid identity with individual claudins (11), with the highest homology in a segment of the first extracellular domain that is thought to bridge the intercellular space (17). In addition, PCLN-1 has a consensus Thr-X-Val PDZ-binding domain at the COOH-terminus (18). Although the intracytoplasmic NH2-termini of the claudins are only six or seven amino acids in length,PCLN-1 encodes a cytoplasmic NH2-terminus of 73 amino acids. This segment is highly hydrophilic with a net positive charge. Interestingly, PCLN-1 encodes a methionine with a suitable Kozak consensus sequence at the position analogous to the start site of other claudins; it is not yet known which of these alternative translation initiation sites is used in vivo. Northern (RNA) blots revealed that PCLN-1 is expressed as a 3.5-kb transcript that is found only in kidney (Fig. 2C) (19). Analysis of genomic organization indicated that PCLN-1consists of five exons, each flanked by canonical splice donor and acceptor sequences (Fig. 2D) (20).

Figure 2

Sequence, structure, and expression of PCLN-1. (A) The deduced amino acid sequence of PCLN-1. Locations of putative transmembrane domains are indicated as TM1 to TM4. Amino acids that are identical to highly conserved residues among claudin family members are highlighted (11, 29). (B) Dendrogram of members of the claudin family, whose relationships were determined using the MegAlign Clustal method (15) (h, human; m, mouse; OSP, oligodendrocyte-specific protein). (C) Northern blot of PCLN-1. An autoradiogram of PCLN-1 hybridized to polyadenylated [poly(A)+] RNA from indicated human tissues is shown; locations of size standards in kilobases are indicated (19). Skel. Mus., skeletal muscle; other tissues showing no detectable expression were spleen, thymus, prostate, testis, ovary, small intestine, colon, and peripheral blood leukocyte. (D) Genomic organization of the PCLN-1 gene. The PCLN-1 protein is encoded in five exons (gray boxes). The codon at the beginning of each exon is indicated, and the base within the codon is denoted by the subscript. (E) Localization of PCLN-1 to specific nephron segments. Rabbit nephron segments were microdissected and tested for the presence of PCLN-1 mRNA by RT-PCR (22). The products were fractionated on agarose gel. P2 and P3, segments of proximal tubule; MTAL and CTAL, medullary and cortical thick ascending limb, respectively; DCT, distal convoluted tubule; CNT, connecting tubule; CCD, cortical collecting duct; OMCD and IMCD, outer and inner medullary collecting duct, respectively.

To determine whether mutations in PCLN-1 cause renal hypomagnesemia, we screened the coding region and intron-exon junctions for variants and then sequenced the variants (21). We identified 10 different mutations that alter the protein in 10 kindreds (Fig. 3). Patients were homozygous for mutations in eight kindreds and compound heterozygotes in two outbred kindreds. The mutations cosegregated with the disease, and none were identified in 160 control chromosomes. Mutations included premature termination as well as splice site and nonconservative missense mutations. For example, affected members of kindred K114 are homozygous for a premature termination codon at position 149, resulting in a protein missing the last three transmembrane domains (Fig. 3A). Similarly, apparently unrelated patients K118-1 and K119-1 both have a homozygous missense mutation substituting arginine for gly cine at codon 239 (Fig. 3B). This mutation introduces a charged residue into the fourth transmembrane domain, replacing the normal glycine residue that is conserved among all members of the claudin family (11, 17). The Gly191 → Arg mutation in K112 (Fig. 3C) alters another completely conserved glycine in the third transmembrane domain.

Figure 3

Mutations in PCLN-1 in recessive renal hypomagnesemia kindreds. PCLN-1 variants were identified by SSCP and sequenced (21). Autoradiograms are shown at the top of each panel. Affected persons are denoted by an asterisk; unrelated normal persons are denoted by N. Variants are indicated by arrows. Below, the corresponding DNA sequences of the sense strand of wild-type (left) and mutant alleles (right) are shown; mutations are indicated by asterisks. Each sequence begins with the first base of a codon, with the exception of (D), which begins in an intron. (A) CGA (R149) to TGA (Term149) mutation. (B) GGA (G239) to AGA (R239) mutation. (C) GGA (G191) to AGA (R191) mutation. (D) G to A mutation changes the first base of exon 4, which is the second base of codon 198, from GGT (G198) to GAT (D198). The mutation is homozygous in K108-1 and heterozygous in K106-1, who is from an outbred kindred and contains another mutation, S235F, on the other allele. (E) Mutations in PCLN-1 in patients with recessive renal hypomagnesemia. Mutations found in patients are indicated in red. The locations of negatively charged amino acids in the first extracellular domain are indicated in yellow (29).

We next determined where PCLN-1 is expressed in the kidney. Using microdissected nephron segments (22), we detectedPCLN-1 mRNA only in TAL and the distal convoluted tubule (DCT) (Fig. 2E). PCLN-1 protein was next localized in kidney sections with the use of an antibody to PCLN-1 (anti–PCLN-1) (Fig. 4A) (23). The protein was detected at intercellular junctions of a subset of renal tubules (Fig. 4, B to D), including all tubule segments staining for the TAL-specific Tamm-Horsfall protein (Fig. 4E). Finally, we performed confocal microscopy using antibodies to both PCLN-1 and occludin, a ubiquitous tight junction protein (24). PCLN-1 and occludin were found to colocalize, indicating that PCLN-1 is a component of the tight junction (Fig. 4, F to O).

Figure 4

Immunolocalization of PCLN-1. (A) Protein immunoblot of bacteria expressing PCLN-1 probed with anti–PCLN-1 (23). A specific band with an apparent molecular mass of 36 kD (arrow) is detected. The remaining panels show immunolocalization of PCLN-1 in frozen human kidney sections. Sections were stained with the indicated antibodies and subjected to confocal microscopy (24). Magnifications, ×630. (B) Three cells in longitudinal section of renal tubule stained with anti–PCLN-1, showing staining of intercellular junctions. (C) Tangential section of tubule stained with anti–PCLN-1, showing staining of intercellular junctions. (D) Cross section of tubule stained with anti–PCLN-1, showing localization to apical intercellular junctions (arrow indicates one junction). Each of these junctions costains with anti-occludin (30). (E) Tangential section of tubule stained with indocarbocyanine (CY3)–labeled anti–Tamm-Horsfall protein (a TAL marker) and fluorescein isothiocyanate (FITC)–labeled anti–PCLN-1. All THP-positive tubules also stain for PCLN-1. The remaining images show colocalization of occludin (FITC-labeled) and PCLN-1 (CY3-labeled) in renal tubules. (F) Composite of 36 confocal images demonstrating colocalization (yellow) of occludin and PCLN-1 in intercellular junctions. (G to I) A single confocal section from (F), showing staining for occludin (G), PCLN-1 (I), and the two superimposed (H), demonstrating colocalization. (Jto L) Individual tubule cell showing staining for occludin (J), PCLN-1 (L), and the two superimposed (K). (M toO) A vertical section through intercellular junctions shown in (J) to (L) composed of 93 confocal sections (in zaxis), showing staining for occludin (M), PCLN-1 (O), and the two superimposed (N), confirming colocalization throughout the junction.

These results identify PCLN-1 as a renal tight junction protein that when mutated causes massive renal magnesium wasting with hypomagnesemia and hypercalciuria, resulting in nephrocalcinosis and renal failure. We infer that these mutations cause loss of normal PCLN-1 function, and that no other genes are redundant in function to PCLN-1.

Despite its critical role in many biological functions, the mechanisms underlying Mg2+ homeostasis have remained obscure. Genetic studies are now providing new insight into homeostatic determinants. In addition to PCLN-1, mutations in the gene encoding the Na+-Cl cotransporter of the DCT have been shown to indirectly cause hypomagnesemia (25). Although the mechanism of Mg2+ resorption in the DCT is uncertain, it is of interest that PCLN-1 mRNA is expressed at this site. In addition, two other genes altering either intestinal or renal Mg2+ handling have been mapped but not yet identified (26).

Our results suggest that PCLN-1 is required for a selective paracellular conductance. We propose that PCLN-1, alone or in partnership with other constituents, forms an intercellular pore permitting paracellular passage of Mg2+ and Ca2+ down their electrochemical gradients. The existence of such pores was predicted from prior physiologic studies characterizing paracellular conductances (27). Such a conclusion is supported by PCLN-1's similarity to claudins, which form intercellular bridges in tight junctions; the location of PCLN-1 in tight junctions of the TAL; the phenotype of patients with mutations in this gene; and the prior physiology implicating the paracellular pathway of the TAL in Mg2+ homeostasis. The hypercalciuria seen in patients with mutations is consistent with physiologic evidence that the paracellular pathway of the TAL mediates resorption of both Mg2+and Ca2+. The lesser dependence of Ca2+homeostasis on this paracellular pathway, as well as the ability of parathyroid hormone to increase intestinal absorption and renal transcellular resorption of Ca2+, may account for patients' maintenance of normal serum Ca2+(28). Alternatively, given physiologic evidence that this paracellular pathway is regulated by Mg2+ concentration, PCLN-1 could be a sensor of Mg2+ concentration that alters a paracellular permeability mediated by other factors. PCLN-1 could conceivably mediate both functions, because tight junctions are in contact with fluid on both sides of epithelial layers. It is noteworthy that the first extracellular domain of PCLN-1 is by far the most negatively charged of the claudin family, with 10 negatively charged residues and a net charge of –5 (Fig. 3E). This high density of negative charges may contribute to the cationic selectivity of this paracellular pathway and could also bind Mg2+ or Ca2+ (or both). Finally, these results raise the possibility that other members of the claudin family also mediate specific paracellular conductances and determine the permeabilities of different epithelia.

  • * These authors contributed equally to this report.

  • To whom correspondence should be addressed. E-mail: richard.lifton{at}


Stay Connected to Science

Navigate This Article