Forkhead transcription factor FoxA1 regulates sweat secretion through Bestrophin 2 anion channel and Na-K-Cl cotransporter 1

Abstract

Body temperature is maintained in a narrow range in mammals, primarily controlled by sweating. In humans, the dynamic thermoregulatory organ, comprised of 2–4 million sweat glands distributed over the body, can secrete up to 4 L of sweat per day, thereby making it possible to withstand high temperatures and endure prolonged physical stress (e.g., long-distance running). The genetic basis for sweat gland function, however, is largely unknown. We find that the forkhead transcription factor, FoxA1, is required to generate mouse sweating capacity. Despite continued sweat gland morphogenesis, ablation of FoxA1 in mice results in absolute anihidrosis (lack of sweating). This inability to sweat is accompanied by down-regulation of the Na-K-Cl cotransporter 1 (Nkcc1) and the Ca²⁺-activated anion channel Bestrophin 2 (Best2), as well as glycoprotein accumulation in gland lumens and ducts. Furthermore, Best2-deficient mice display comparable anhidrosis and glycoprotein accumulation. These findings link earlier observations that both sodium/potassium/chloride exchange and Ca²⁺ are required for sweat production. FoxA1 is inferred to regulate two corresponding features of sweat secretion. One feature, via Best2, catalyzes a bicarbonate gradient that could help to drive calcium-associated ionic transport; the other, requiring Nkcc1, facilitates monovalent ion exchange into sweat. These mechanistic components can be pharmaceutical targets to defend against hyperthermia and alleviate defective thermoregulation in the elderly, and may provide a model relevant to more complex secretory processes.

Sweat glands provide the major organ cooling down body temperature in humans exposed to a hot environment, physical exercise, or fever. Each sweat gland is a single tube consisting of functionally distinctive duct and secretory portions. A sweat duct comprises a helix-shaped epidermal duct, a straight dermal duct, and a short coiled extension of the dermal duct continues into a coiled secretory portion (Fig. 1C). The secretory portion, primed to respond to stimulation from sympathetic nerves, includes “clear” and “dark” secretory cells and supportive myoepithelial cells (1).

Fig. 1.

FoxA1 knockout causes anhidrosis in mice with accumulation of glycoproteins in sweat glands. (A) Iodine-starch sweat test reveals complete anhidrosis in FoxA1-mu footpads. Purple dots in control mice represent sweating spots. (B) H&E staining shows smaller but fully formed sweat glands in FoxA1-mu mice (Upper). Higher magnification (Lower) shows the blockage of secretory lumens by eosinophilic amorphous substances (arrows) in the mutant sweat glands, compared with fully open control lumens. Magnification, 400× for upper panels and 1,000× for lower panels. (C) Epidermal (Upper) and coiled (Lower) ducts were also filled with eosinophilic material in the mutant mice. At right, diagram of sweat gland structure. (D) Eosinophilic material in the epidermal ducts and secretory lumens of FoxA1-mu mice is PAS⁺ and α-amylase resistant (arrows). Magnification, 1,000×.

The clear and dark cells secrete primary sweat that is isotonic, similar to plasma (2). The ionic content of primary sweat is then partially reabsorbed, mainly in the coiled sweat duct, resulting in hypotonic final sweat (2). The dynamic process of sweat secretion is thought to involve two primary forms of transport. Na-K-Cl cotransport is thought to play an imperative role in mediating sweat secretion. In addition, less-studied Ca²⁺-dependent transport, largely separated from the Na-K-Cl system, is absolutely required for sweat secretion. To date, however, no model has integrated these two features.

First molecular insights into the sweating apparatus came from the study of anihidrotic ectodermal dysplasia (EDA), a disorder that is caused by mutations in Eda pathway genes (3), most often in ectodysplasin, the product of the EDA gene (4). Sweat glands develop in a cascade pathway initiated by Wnt/Eda and involving intermediate stages regulated by factors that include the secreted morphogen sonic hedgehog. A hint to the unique mechanism of sweating was the finding of high expression of the forkhead transcription factor, FoxA1, in maturing mouse sweat glands but not in hair follicles or skin epidermis (5). FoxA1 is known to aid in development of endoderm-derived organs (6–10), and was shown to be involved in water reabsorption in mesoderm-derived kidney (11). Its unexpected expression in the ectoderm-derived sweat glands prompted us to study its possible role there. Here we show that FoxA1 is a master regulator of sweat secretion, working through the Best2 anion channel and Nkcc1 ion cotransporter.

Results

Skin-Specific FoxA1-Deficient Mice Show Absolute Anhidrosis with Striking Accumulation of Periodic Acid-Schiff–Positive Glycoproteins in Sweat Gland Lumens and Ducts.

To study FoxA1 function in sweat glands, we generated skin-specific FoxA1-deficient (FoxA1-mu) mice by crossing FoxA1^loxP/loxP mice (12) with transgenic mice bearing Cre driven by a Krt14 promoter. The resultant mutant mice lack the critical FoxA1 exon 2 (forkhead domain) specifically in the skin (Fig. S1A). The FoxA1-mu mice were grossly similar to control littermates. However, iodine-starch sweat tests revealed striking anhidrosis (Fig. 1A). Histological analysis showed anatomically fully formed sweat glands (Fig. 1B), with lumens and all ducts, including epidermal ducts in the outermost stratum corneum layer of skin, were entirely blocked by amorphous eosinophilic protein (Fig. 1 B and C). Time-course studies revealed that morphological alterations were detectable in mutant mice as early as postnatal day 10 (P10) (Fig. S1B). Eosinophilic protein was discernible in lumens by P8 in both control and mutant mice, but disappeared in the control mice concomitant with lumen opening at P10. In the FoxA1-mu mice, eosinophilic proteins persisted (Fig. S1B).

The eosinophilic material proved to contain the periodic acid-Schiff (PAS)–positive and diastase-resistant sweat glycoproteins (Fig. 1D), which are produced by dark secretory cells (compare with Fig. 2, Left) (13).

Fig. 2.

FoxA1 is expressed in the nuclei of dark cells. (Left) Diagram of cell arrangement in cross-section of a sweat gland secretory portion. (A and B) FoxA1 is expressed in the S100⁻, CAII⁻ secretory cells in control sweat glands. (C) FoxA1 is expressed in CGRP⁺ secretory cells. Dotted circles indicate secretory portion.

FoxA1 Is Expressed in the Dark Secretory Cells.

By coimmunostaining with Krt8 and Krt14, markers for sweat-gland secretory cells and myoepithelial cells, we found FoxA1 to be expressed in the nuclei of scattered secretory cells (Fig. S1C). In a control experiment, we confirmed that FoxA1 was not expressed in hair follicles or skin epidermis (Fig. S1D). Coimmunostaining with S100 and carbonic anhydrase II (CA II), clear cell markers, and the calcitonin gene-related peptide (CGRP), highly expressed in the dark cells of human sweat glands (14, 15), revealed exclusive expression of FoxA1 in the S100⁻, CA II⁻, CGRP⁺ secretory cells (Fig. 2). Thus, FoxA1 is inferred to be expressed in dark cells. FoxA1-mu sweat glands expressed all five cell markers analyzed, consistent with histological findings of fully formed sweat glands in these mice (Fig. 2 and Fig. S1C). These findings suggested that FoxA1⁺ dark cells are primary instigators of sweat secretion.

Expression Profiling Revealed Sharp Down-Regulation of Best2 in the FoxA1-mu Sweat Glands.

To characterize the role of FoxA1 on a molecular level, we assessed its downstream effector genes in sweat glands. Expression profiling was done on footpad skin, the only site of sweat glands in mice. Footpads were sampled from control and FoxA1-mu mice at P10, P14, and P31, representing the normal times of lumen opening, completion of gland development, and steady-state function. In addition to FoxA1, only expression of Best2, 9930023K05Rik, and Nkcc1 was significantly affected by microarray and quantitative RT-PCR (qRT-PCR) in FoxA1-mu mice at all three time-points (Fig. 3A and Fig. S2 A and C). Best2, a Ca²⁺-activated anion channel, and Nkcc1, encoding the Na-K-Cl cotransporter, were particularly suggestive for an involvement in sweating, and were studied further.

Fig. 3.

Dramatic down-regulation of Best2 in FoxA1-mu sweat glands. (A) Scatter plots shows number of significantly affected genes in microarray analysis. Red and green dots represent significantly down- or up-regulated genes in FoxA1-mu sweat glands, respectively. (B) qPCR assay; FoxA1 expression was confirmed undetectable in FoxA1-mu sweat glands. (C) qPCR assay; Best2 expression is at background in FoxA1-mu sweat glands. (D) Best2 protein is undetectable in FoxA1-mu sweat gland protein analyzed by Western blotting. (Upper) Arrow indicates FoxA1 position. (Lower) Loading control (Coomassie blue staining). (E) Immunofluorescent staining shows Best2 expressed in the basolateral and apical membranes of FoxA1⁺ secretory cells in control sweat glands (Left), but not in FoxA1-mu sweat glands (Right).

As expected, Eda, upstream of FoxA1, was unchanged in its expression in FoxA1-mu sweat glands by qRT-PCR analysis (Fig. S2B), whereas FoxA1 expression was at background (Fig. 3B). Best2, the gene most strikingly affected among the three consistently altered (Fig. S2A), was down-regulated more than 10-fold (Fig. 3C). Best2 is of particular interest because it was recently shown to regulate bicarbonate transport in the colon (16), and failure of bicarbonate transport was the only defect found in Best^−/− mice. Although not previously reported in skin, Best2 has also been found in parotid gland (17) and ciliary body in the eye (18), tissues also involved in secretory processes. By immunostaining, we found Best2 expression in control sweat glands, exclusively in the basolateral and apical membranes of FoxA1⁺ secretory cells (Fig. 3E). In contrast, both in Western blot assay (Fig. 3D) and immunostaining (Fig. 3E), it was totally undetectable in FoxA1-mu footpads.

Best2^−/−-LacZ Mice Display Essentially Identical Sweat-Gland Phenotypes with FoxA1-mu Mice.

If Best2 is a critical target of FoxA1 in the sweat glands, it should mediate all or part of FoxA1 function. We therefore examined Best2^−/−-LacZ mice, in which the Best2 gene is replaced by LacZ (18). These mice indeed showed highly defective sweating in sweat tests, ranging from severe to complete anhidrosis (Fig. 4A). In qRT-PCR assays, we found that Best2 mRNA was essentially gone in Best2^−/−-LacZ sweat glands (Fig. 4B). As in FoxA1-mu mice, sweat glands were fully formed in the Best2^−/−-LacZ mice, although lumens and ducts were much narrower in the mutant mice (Fig. 4C). Eosinophilic, PAS-staining material again accumulated in the lumens and ducts (arrows in Fig. 4 C–E), although less severely than in FoxA1-mu mice.

Fig. 4.

Best2^−/− mice show sweat gland phenotypes like FoxA1-mu mice. (A) Sweat test: anhidrosis in Best2^−/− mice, similar to FoxA1-mu (compare with Fig. 1). (B) qPCR: Best2 expression is at background level in Best2^−/− sweat glands. (C) Sweat glands are fully formed in the Best2^−/− mice, but lumens are much smaller than in controls (Upper). Eosinophilic material in coiled ducts in Best2^−/− mice (higher magnification with arrows, Lower). Magnification, 400× for upper panels and 1,000× for lower panels. (D) Eosinophilic material also in the stratum corneum epidermal ducts of Best2^−/− mice (arrows). Magnification, 1,000×. (E) PAS staining: accumulation of α-amylase resistant glycoprotein in epidermal ducts and secretory lumens of Best2^−/− mice (arrows). Magnification, 1,000×. (F) Immunofluorescent staining: normal expression pattern of FoxA1 protein in Best2^−/− sweat glands.

Through X-Gal staining, Best2 promoter-driven β-galactosidase was found to be active in a group of scattered secretory cells in the Best2^−/−-LacZ mice (Fig. S3A). In immunostaining, Best2 was at background levels in Best2^−/−-LacZ sweat glands, but FoxA1 was normally expressed, consistent with its regulatory position upstream of Best2 (Fig. 4F). Furthermore, β-galactosidase was exclusively expressed in the cytosol of FoxA1⁺ secretory cells in the Best2^−/−-LacZ mice (Fig. S3B), in agreement with Best2 protein localization in control sweat glands (Fig. 4F). Best2 is thus a primary effector responsible for the anhidrosis caused by ablation of FoxA1.

Nkcc1 Is Significantly Down-Regulated in the FoxA1-mu Sweat Glands.

Nkcc1, a major effector in the current ionic Na-K-Cl cotransporter model for cholinergic sweat secretion stimulated by muscarinic agents in vitro (2), is expressed in basolateral membranes of secretory cells (19). By immunostaining in control sweat glands with two different antibodies against N- and C-terminal Nkcc1 epitopes, we confirmed that all FoxA1⁺ and some FoxA1⁻ secretory cells express Nkcc1 in their basolateral membranes (Fig. S3C). In contrast, FoxA1-mu sweat glands showed down-regulation of Nkcc1 mRNA about twofold by microarray and qRT-PCR (Fig. S2 A and C), and its protein level, assayed by immunostaining, was sharply reduced (Fig. S3C). The weak staining was seen in all secretory cells, perhaps indicating a general posttranscriptional dependence of Nkcc1 protein levels on FoxA1.

As expected, Nkcc1, like FoxA1, was normally expressed in Best2^−/−-LacZ sweat glands by microarray and qRT-PCR assays (Fig. S3D). As in control sweat glands (Fig. S3C), Nkcc1 immunostaining was seen in basolateral membranes of both secretory cells expressing and those not expressing FoxA1. Therefore, regulation of Nkcc1 levels by FoxA1 is independent of Best2 action.

Discussion

Sweat glands, as the smallest but robust secretory organ, has attracted study of its operating mechanism for more than half a century (2). The earliest leak-pump model was replaced in the 1970s by the Na-K-Cl cotransport model, now widely accepted for the predominant cholinergic sweat secretion (1, 2). However, the mechanism, especially the genetic regulation of sweating, has been largely unknown. In this study, we found that a FoxA1-Best2 cascade, along with the Nkcc1 cotransporter, is central for sweat secretion, with FoxA1 as the first transcription factor linked to the regulation of sweating.

Two striking findings of the present study were anhidrosis and excessive accumulation of glycoproteins in the FoxA1-mu and Best2^−/− sweat glands. The primary effect of these genes in sweat glands should explain both these phenotypes. Unlike the features of Miliaria Rubra, a human disorder caused by blockage of sweat ducts by glycoproteins or keratin debris (20), neither FoxA1-mu nor Best2^−/− mice showed any indication of edema or inflammation in sweat glands. In accordance with that view, expression profiling showed no alteration of mucin gene expression. Thus, it is more likely that mucinous glycoproteins, which are produced by dark cells in small quantities and normally washed out by sweating, accumulate in FoxA1-mu or Best2^−/− sweat glands secondary to anihidrosis.

Concerning the site of critical events, FoxA1 regulation of sweat secretion implies that although clear cells have been credited with active sweating (1, 2), dark cells expressing Best2 play a previously unrecognized and essential role. Similarly, Best2 in the colon was initially expected to be expressed in enterocytes, but was instead found to be expressed in the mucin-producing goblet cells (21). It was previously suggested that interactions between dark and clear cells are required to maintain their identity, as they lose their morphological characteristics rapidly when separated in culture (2). Consistent with interdependency of the two types of cells, the Nkcc1 cotransporter was normally expressed in anhidrotic Best2^−/− sweat glands, demonstrating that Nkcc1 or other ion channels expressed in clear cells are ineffective in generating sweat in the absence of Best2 in dark cells. However, detailed steps whereby dark cells may regulate further action in clear cells remain to be determined. In cases which may be analogous to sweat glands and colon, the ciliary body of the eye and the salivary gland, both secretory organs, consist of two types of epithelia, one of which contains Best2 (17, 21, 22). It will be intriguing to test whether Best2 is regulated by FoxA1 in these organs, and how the two types of epithelia cooperate through Best2 for fluid secretion.

The process of sweat secretion is multifaceted, involving central and sympathetic nervous system regulation, clear and dark secretory cells, and ion and water transport, along with ion reabsorption by Na pump, cystic fibrosis transmembrane conductance regulator, and carbonic anhydrase 12 (1, 2, 23, 24). The model for sweat secretion remains incomplete, but the findings suggest some features of an overall model. From previous studies, Na⁺, K⁺, and Cl⁻ are the major ionic components of sweat in the accepted Na-K-Cl cotransporter model for cholinergic sweat secretion (1, 2). In this model, cholinergic input opens K⁺ and Cl⁻ channels, and the resulting chemical potential from KCl efflux then activates the Nkcc1 cotransporter (1). As for the Ca²⁺-dependent model involved in both cholinergic and α-adrenergic sweat secretion, sweating is abolished when Ca²⁺ is removed, and conversely, Ca²⁺ influx into the cytoplasm of secretory cells induces sweat secretion in most (but not all) conditions (1, 2). It was suggested that Ca²⁺ influx may contribute to the opening of K⁺ and Cl⁻ channels by an unknown mechanism (2), and thus could also be involved in subsequent activation of Nkcc1. Two transport routes may thus have some crosstalk at K⁺ and Cl⁻ channels in the early steps of sweat secretion (2). Best2 is itself a Ca²⁺-activated anion channel (25, 26), and Best2 catalyzes bicarbonate transport as part of ion flux in the colon (21). It remains to be seen whether Ca²⁺-activated Best2 in dark cells regulates the opening/closing of K⁺ or Cl⁻ channels in clear and dark cells, involving a bicarbonate gradient, and further effecting subsequent activation of Nkcc1. Identification of FoxA1 as a master regulator integrates transport by Nkcc1 with the indispensable requirement for Ca²⁺ activation of Best2 channels to cooperate in sweat secretion (27).

The mode of thermoregulation is drastically transformed during evolution. In most mammals, the hair coat helps to modulate temperature, and sweat glands are located only in hairless footpads. However, some primates and Homo sapiens have much less hair, and sweat glands extensively spread over the entire body become the substantive thermoregulatory organ. Sweat glands are efficient because they actively dissipate heat; and as a result humans, unlike other animals, can run long-distance races and even super marathons. [Horses expand another type of sweat gland (apocrine) to sustain comparable cooling and running capability.] Notably, FoxA1 in mouse skin is exclusively expressed in sweat glands, but not in hair follicles or skin epidermis (Fig. S1D), and we infer that FoxA1 is part of a selective evolutionary option to implement human body temperature regulation by sweating. Thus far, one patient with idiopathic anhidrosis showed pathophysiology similar to that seen in FoxA1 or Best2 mutant mice (28), and it remains to be determined if variation in either of the genes is implicated in such patients or in differential individual rates of sweating.

Materials and Methods

Generation of Animal Models.

All animal study protocols were approved by the National Institute on Aging Animal Care and Use Committee. Homozygous FoxA1^loxP/loxP mice (12), harboring a floxed exon II of mFoxA1 gene, were crossed with K14-Cre transgenic mice (Jackson Laboratory) to yield heterozygous FoxA1^+/loxP-K14-Cre mice. Heterozygous mice were crossed with each other to generate skin-specific FoxA1-deficient mice (FoxA1-mu). Genotyping was done by PCR, as shown in Fig. S1A. Primer set for FoxA1 genotyping: forward, 5′-CTGTGGATTATGTTCCTGATC; reverse, 5′-GTGTCAGGATGCCTATCTGGT. Generation of Best2-deficient mice, in which exon2 and parts of exon1 and 3 were replaced by a LacZ, was reported previously (18). Primer set for Best2 genotyping: Forward for Best2, 5′-ACGTGACTCTGGTGGTGCAT; reverse for Best2: 5′-CAGGGCACCCAGTATTTGTT; forward for LacZ-Neo: 5′-GCCTGAAGAACGAGATCAGC.

Sweat Test and General Histology.

The iodine-starch sweat test was done as previously reported (29). Briefly, a hind paw of a wild-type or mutant mouse was painted with iodine/alcohol solution (1 g iodine/50 mL ethanol). Once dry, the surface was painted again with starch-oil (10 g starch/10 mL castor oil). Purple sweating spots start to form 2–3 min later and peak at around 10 min.

For general histology and immunofluorescent or PAS staining, footpads and hairy back skin were taken from mice, fixed in 10% formaldehyde solution (Ricca Chemical) overnight, dehydrated, and embedded in paraffin. Five-micrometer paraffin sections were cut, and deparaffinized, and H&E staining was carried out with the manufacturer's standard protocol (Sigma).

Immunofluorescent Staining, Western Blotting, PAS Staining, and X-Gal Staining.

For immunofluorescent staining, antibodies against β-galactosidase (Abcam; ab616, 1:300 dilution), Best2 (Santa Cruz; sc-98568, 1:1,000), CAII (Abcam; ab6621, 1:500), CGRP (Abcam; ab36001, 1:300), FoxA1 (Abcam; ab23738, 1:200; Santa Cruz; sc-6553, 1:200), Krt8 (Progen; GP-K8, 1:200), Krt14 (Progen; GP-CK14, 1:200; Covance; PRB-155P, 1:1,000), Nkcc1 (Abcam; ab104693, 1:1,000; ab59791, 1:500), and S100 (Abcam; ab4066, 1:100) were incubated with deparaffinized, “antigen-unmasked” (antigen unmasking solution; Vector Laboratories, H3300) sweat gland sections overnight at 4 °C. AlexaFlour conjugated secondary antibodies were used to detect primary antibodies (Invitrogen), and images were taken with a Deltavision Restoration Microscope System (Applied Precision).

For Western blotting, proteins were extracted from frozen footpad skin of control and FoxA1-mu mice by homogenation in RIPA buffer first (Sigma; R0278) (the soluble fraction), and the pellets were re-extracted by RIPA+1%SDS (the insoluble fraction) (30). The reextracted “insoluble” proteins were transferred to a nylon membrane after SDS gel electrophoresis, and signals were detected by ECL (Amersham), after incubation with primary (anti-Best2 antibody) and HRP-conjugated secondary antibodies.

PAS staining was done with paraffin sections from control, FoxA1-mu, and Best2^−/− sweat glands with a PAS staining kit with a microwave procedure (Sigma; 395B). To detect diastase resistance of glycoproteins, α-amylase solution (0.5%) was applied to sections after deparaffinization, and followed by PAS staining. Sections were counterstained with hematoxylin as the final step.

The procedure of X-Gal staining for LacZ expression was reported previously (18). Briefly, 10-μm frozen sections of footpad skin were fixed in 4% paraformaldehyde and incubated overnight in a 1 mg/mL solution of X-Gal (Tissue Stain Solution; Chemicon). After washing, the sections were counterstained with nuclear fast red.

Expression Profiling and qRT-PCR.

For expression profiling of FoxA1-mu sweat glands, hairless fore-footpad skin was collected from FoxA1-mu and control littermates at P10, P14, and P31. Three skin samples from three mice for each genotype at each time point were used for biological replicates. Total RNAs were isolated with TRIzol (Invitrogen), precipitated by 7.5M LiCl (Ambion), and cyanine-3-labeled cRNAs were hybridized to the NIA Mouse 44K Microarray v3.0 (Agilent Technologies) (30). Triplicate data were analyzed by ANOVA. Genes with false-discovery rate < 0.05, fold-difference > 1.5, and mean log intensity > 2.0 were considered to be significant (30). Hybridization data were deposited in the Gene Expression Omnibus (GEO) with accession number GSE32347. One-step TaqMan qRT-PCR was carried out with ready-to-use probe/primer sets for FoxA1, Best2, Eda, and Nkcc1 (Applied Biosystems).