In this study we begin to address the question of whether members of the claudin family confer ionic selectivity to the paracellular pathway. We observed that overexpression of claudin-4 in MDCK cells decreases transmonolayer conductance by decreasing paracellular Na+
permeability. There is no effect on the permeability for Cl–
or flux for a noncharged solute. The close correlation between the level of claudin-4 expression and change in both conductance and ionic selectivity suggest it is directly responsible for these changes. These results are consistent with a model in which claudin-4 forms channels through the tight junction that discriminate against Na+
ions and are indifferent to Cl–
ions. The results of bi-ionic potential studies imply that the permeability was also decreased proportionately for the other alkali metal cations, K+
, and Cs+
. Although recent studies by Inai et al. (12
) and McCarthy et al. (13
) showed that overexpression of claudin-1 increased TER in MDCK cells, the present data are the first direct evidence, to our knowledge, that overexpression of a claudin can confer paracellular ionic discrimination. Previous studies have documented the varied distribution of individual claudins (5
) and led to the speculation they might be responsible for the variable electrical properties of tight junctions in different tissues. Demonstration of the ability of claudin-4 to alter Na+
selectivity supports this model and is an important first step in understanding underlying mechanisms responsible for tissue-specific tight-junction properties.
Several lines of evidence strongly suggest claudin-4 is affecting the paracellular pathway. First, paracellular conductance greatly exceeds transcellular conductance in MDCK cells, therefore if claudin-4 increases TER it must do this through an effect on tight junctions (20
). Second, the number of junction strands is increased by overexpression of claudin-4, while the levels of several other strand proteins are unaffected. Thus we presume the additional strands are made predominately of claudin-4 in a position to influence the barrier characteristics of the junction. This does not rule out the possibility that claudin on the lateral cell surface, not assembled into strands, can influence paracellular permeability, and this could be tested experimentally by using osmotic gradients to shrink or swell the lateral intercellular space. Third, the instantaneous current-voltage relationships were linear and symmetrical in both uninduced and induced MDCK cells. Transmembrane carriers would be expected to have a limited capacity to conduct current and result in a nonlinear I-V relationship. The observed lack of saturation is consistent with the major avenue of conductance being between the cells rather than through membrane carriers. Finally, amiloride, bumetanide, and SITS were used to block major transcellular ion-transporting channels. Since dilution potentials in both uninduced and induced cells were unaffected by the presence of these inhibitors, it is likely that the potentials were generated by the paracellular rather than transcellular movement of Na+
It should be noted that we chose to ignore any contribution to the dilution potential from ions other than Na+ and Cl–. This is commonly done and valid to a first approximation since the other ions are present at lower concentrations and their transmonolayer ratios were kept constant during imposition of the NaCl gradient. It is possible that the other ions compete with Na+ and Cl– differently at the two NaCl concentrations. This anomalous behavior would be expected to contribute equally to the dilution potential in both uninduced and induced cells. Thus, the increase in PCl/PNa observed with overexpression of claudin-4 must reveal a true change in the relative selectivity even if the exact magnitudes are approximations.
Overexpressing claudin-4 does not stimulate an overall increase in the levels of other tight-junction components, consistent with a previous study by McCarthy et al. on claudin-1 (13
). Notably, ZO-1 levels do not change in the face of claudin-4 induction. One postulated role for the ZO proteins is to scaffold the integral membrane proteins and link them to the actin cytoskeleton (17
). The physiologic effects of a large increase in claudin-4 without an increase in ZO-1 suggest that either this linkage is not required for barrier function or that the ZO proteins or other components involved in the linkage are not limiting. That levels of occludin and the other claudins measured do not diminish in the face of claudin-4 overexpression suggests that claudin-4 does not act to replace these integral membrane tight-junction proteins, but adds to them. Since the composition of the tight junction particle is unknown, it is impossible to predict how overexpression of claudin-4 alters the stoichiometry of claudins and occludin in the new tight-junction strands.
Many studies have now shown dissociation between changes in the barrier properties of electrical conductance for ions and the flux for uncharged solutes (2
). Transmonolayer conductance is an instantaneous measure of ion permeability across the paracellular pathway during imposition of an electrical field, reflecting only the number and nature of channels at that instant. This property varies over five orders of magnitude among normal tissues, yet their size selectivity for noncharged solutes is very similar (1
). This supports the idea that the channels discriminate on the basis of charge but are of the same size. Mannitol is a small, uncharged solute commonly used to probe the size of the paracellular channels; the flux for mannitol is normally measured over minutes to hours. Although conductance and flux often change together when tight-junction function is compromised, there is considerable support for the idea that the barriers measured by these two methods are functionally and physically distinct. The longer times over which flux is measured implies that if strands were labile, i.e., separated and reformed over time, then the paracellular route available for diffusion would be different from that measured by electrical conductance and not dependent on claudin-based channels. A number of recent studies document increased TER (or decreased conductance) with either no effect on or seemingly paradoxical increases in mannitol flux. Examples include EGF treatment on LLC-PK1
), overexpression of occludin (31
), overexpression of claudin-1 (13
), and RhoA activation in MDCK cells (33
). The observation that overexpressing claudin-4 decreases ionic conductance without altering the rate of mannitol flux is consistent with the idea that the selectivity but not the size of the paracellular channels has been altered. Although a number of models have been proposed, reasons for the lack of correlation between the paracellular movement of ions, water, and small nonionic solutes like mannitol are still unclear (2
The ion selectivity of MDCK tight junctions was first determined nearly 25 years ago (20
). In those studies, MDCK cells were found to be cation selective and the rank order of cation permeability was determined, but there were some differences among the studies. Our results in uninduced tet-off MDCK cells corresponded well with those published by Rabito et al. (26
), although our cells were slightly less cation selective (PNa
Cl = 2.5 in our cells compared with PNa
Cl = 4.4). Induction of claudin-4 resulted in a specific decrease in absolute sodium permeability and changed the PNa
to 1.6. Although our study is the first report in which a change in ionic selectivity can be attributed to overexpression of a specific tight-junction protein, there are previous reports of experimental treatments that alter the ionic selectivity of the tight junction. These include the effects of EGF on LLC-PK1
) and TNF-α (35
) on CACO-2 cells, among others. One possible explanation is that these treatments alter claudin expression profiles in the tight junction, with concomitant changes in paracellular permeability characteristics. This form of regulation of paracellular permeability would then also be expected to exist in vivo.
The molecular basis by which claudin-4 influences paracellular selectivity remains unknown. A likely possibility is that claudins form selective channels in the tight junction through their oligomerization into particles and lateral cell-cell contacts of the particles. The variable amino acid sequences of their extracellular domains, particularly charged side chains, might be positioned to modify the selectivity of the channels (35
). Consistent with this model is the observation that the extracellular sequences of claudin-16/paracellin are among the most acidic. Indirect evidence suggests claudin-16 permits the paracellular passage of Mg++
and other cations (6
). In preliminary studies, we observe that overexpressing claudin-2 in MDCK cells results in an increase in paracellular conductance resulting from an increase in PNa+
. This behavior is opposite that associated with overexpression of claudin-4, again supporting the idea that each claudin contributes a different channel-like property. Since it has been demonstrated that some combinations of various claudins can interact both within strands and across cells (36
), the combinatorial possibilities offered by all 20 claudins could easily provide extensive physiologic variability among tissues. Until we better understand the nature of the tight-junction particle and how it combines with partners on the opposing cell as well as with plaque proteins, our interpretation of the electrophysiologic data will be necessarily limited.