Glucocorticoids and myosin5b loss of function induce heightened PKA signaling in addition to membrane traffic defects

Two intestinal cell lines deficient in Myo5b display changes in PKA-dependent phosphorylation patterns that mimic MVID: effect of dexamethasone

CaCo2 cells (Muller et al., 2008) and the subclone CaCo2BBe (C2BBe; Peterson and Mooseker, 1992), closely resemble villus enterocytes of the small intestine ( Collaco et al., 2010; Jakab et al., 2011; Knowles et al., 2014). Effective (>90%) silencing of Myo5b with lentivirus-delivered short hairpin (shRNA) constructs was regularly achieved (Figure 1A) after stringent selection in puromycin (Kravtsov et al., 2016). Under these conditions, C2BBe cells display all the morphologic features of human MVID enterocytes including the loss of microvilli and the appearance of MI (Kravtsov et al., 2014; Vogel et al., 2015), which can be visualized as a steep decrease in apical phalloidin signal and intracellular phalloidin-positive inclusions in a fraction of the cells (Figure 1B, arrow).

To determine whether the activity of PKA is specifically exerted in the apical domain of epithelial cells, where CFTR molecules are localized, we used a widely validated antibody recognizing phosphorylated PKA consensus target sites. This antibody has been used in immunohistochemistry, immunofluorescence, and immunoblot assays in more than 40 publications. We validated this antibody by immunoblot of forskolin (cAMP)-stimulated C2BBe cell extracts and with a PKA inhibitor (Figure 1C and Supplemental Figure 1A). This antibody will be referred to as PKA p-substrate antibody hereafter. As expected, the antibody recognized several bands. Some bands appeared only in the forskolin-treated cells, while other bands increased in intensity under forskolin stimulation (Figure 1C). Because localization of PKA activity is highly regulated, we wanted to understand the distribution of PKA-phosphorylated targets in intestinal cells. C2BBe cells transduced with a scrambled RNA-expressing lentivirus (scr) showed mostly lateral membrane localization of PKA-phosphorylated proteins (Figure 1D, arrows) with many cells lacking apical signal. Myo5b knockdown (kd) or dexamethasone treatment of scr cells changed this distribution to a broadly cytoplasmic signal but did not result in a clear apical localization. However, in Myo5b kd cells previously treated with dexamethasone the signal was concentrated and enhanced at the apical domain (Figure 1D, bottom panel). Semiquantitative measurements of pixel intensities showed significantly increased signal in the apical domain and decreased signal in the basolateral region and cytoplasm (Supplemental Figure 1B).

Another epithelial intestinal cell line, T84, resembles intestinal crypt cells, and has been shown to closely mimic the electrophysiologic properties of secretory intestinal epithelia (Tousson et al., 1996; Silvis et al., 2009). These cells form tight junctions (Supplemental Figure 2A), albeit their differentiation is slower than C2BBe. They can also be grown in spheroids, which, after 2 wk in Matrigel, show mostly single-layered cysts, with only sporadic double-layered cells (Supplemental Figure 2B, arrows). Importantly, the cells appear polarized, but the lumens are usually collapsed, as expected from cells with a net absorptive fluid transport (Supplemental Figure 2B). We previously used electrophysiologic approaches to demonstrate stimulated CFTR ion transport activity in Myo5b-deficient T84 cells (Kravtsov et al., 2016). Accordingly, these cells were used extensively in functional studies here (next section). In addition, we used T84 cells in the current studies as a second cell line to decrease the chance that the observations may be due to random changes in the genetic background or signaling related to transformation. The same lentivirus-delivered anti-Myo5b shRNA was used in T84 cells, resulting in a similar steep down-regulation of Myo5b (Figure 2C). Pixel intensities were quantified in regions of interest (ROIs) randomly localized at the apical domain by grid overlapping (example shown in the leftmost panel of Figure 2A, and the confocal level and depth of ROIs in Figure 2B). Unlike C2BBe cells, only apical PKA p-substrate signal was observed in T84 cells (Figure 2B). PKA-phosphorylated target showed minimal changes when scr cells were compared with Myo5b kd cells in the absence of dexamethasone (Figure 2, A and D). However, PKA substrate signal increased 2.5-fold when the cells were incubated with forskolin, regardless of Myo5b expression indicating that PKA is available and in active conformation in kd or scr controls. Importantly, average pixel intensities significantly increased threefold in Myo5b-deficient cells incubated in the presence of dexamethasone, a level indistinguishable from the forskolin-positive control (Figure 2, A and D).

Myo5b-deficient intestinal epithelial cells show apical fluid and Cl⁻ secretion in the presence of dexamethasone that can be reversed by CFTR inhibitors

Fluid transport was measured in cells grown on filters by a gravimetric method as described before for C2BBe cells (Kravtsov et al., 2014). These experiments were performed in T84 cells, because these cells express CFTR and display electrophysiologic features similar to those in the intestinal epithelium with Cl⁻ secretion more robust than other epithelial models including CaCo2 cells (Dharmsathaphorn et al., 1984; Resta-Lenert and Barrett, 2002). The gravimetric method showed fluxes of similar magnitude to those obtained by others with different approaches (Toriano et al., 2001). The spontaneous absorptive fluid flux average in T84 cells was 1 µl/h/cm² (mucosal to serosal; Figure 3A). As a comparison, the absorptive fluxes reported in C2BBe were 3.1 µl/h per cm² (Kravtsov et al., 2014). This is consistent with crypt-like T84 cells as compared with villus-resembling absorptive C2BBe cells. As in the case of C2BBe, Myo5b kd decreased the absorptive flux, but did not revert it into a net secretory (serosal-to-mucosal) fluid flux. As a positive control for secretory response, forskolin induced a significant net fluid secretion in both control and Myo5b kd cells. The forskolin-­induced response was decreased in Myo5b kd cells but not abolished (Figure 3A). This is expected for cells expressing only a fraction of the normal CFTR on the apical surface. These findings are consistent with previous electrophysiologic measurements of forskolin-stimulated CFTR anion currents in T84 cells deficient in Myo5b (Kravtsov et al., 2016). They are also consistent with decreased swelling of forskolin-stimulated intestinal organoids generated from mice deficient in Myo5b grown in low GC levels (Schneeberger et al., 2015). However, when the cells were grown in dexamethasone, the unstimulated Myo5b kd monolayers showed spontaneous fluid secretion at 27% of the maximum secretory flux in forskolin stimulation in KD cells, or 11% of the flux in forskolin-stimulated scr cells (Figure 3A, green arrow). Importantly, this serosal-to-mucosal net fluid flux was abolished by the specific CFTR inhibitor CFTR172 and was independent of secretagogues. Moreover, our finding of a net secretory flux of 0.8 µl/h per cm² in the Myo5b kd and dexamethasone condition may seem very small. However, considering that the secretory surface of the newborn small intestine is ∼25,000 cm² (Struijs et al., 2009), it is consistent with 480 ml fluid secretion per day. That is well within the volume range of diarrhea of a 3 kg MVID patient (Cutz et al., 1989).

To independently corroborate fluid secretion, we used spheroid cultures of T84 cells in Matrigel. The diameter of the cysts grown for 10 d did not significantly change in a 23-h period, either in scr- or Myo5b kd cells (Figure 3B). Likewise, when scr T84 cells were supplemented with dexamethasone, the diameter change was <20% (Figure 3B and Supplemental Video_­control). However, in Myo5b kd cultures supplemented with dexamethasone, the cysts significantly grew 62% (Figure 3B and Supplemental Video_dexa). These results support the hypothesis that Myo5b-defective cells secrete fluid in the presence of postnatal physiological levels of GCs.

To determine whether fluid secretion is due to CFTR-dependent apical Cl⁻ extrusion in epithelial cells deficient in Myo5b, standard electrophysiology experiments in an Ussing chamber were performed. Transepithelial electric current (Isc) is the result of a subset of cellular ion transport mechanisms which are electrogenic. CFTR-mediated Cl⁻ secretion is among them (Kato and Romero, 2011). First, T84 cells in tissue culture medium DMEM, were challenged with forskolin from the apical side. In scrambled shRNA-expressing cells, stimulation caused a robust increase in Isc, with a transient peak in the range 40–100 µA/cm², which decreased to a steady-state plateau after 15–20 min (Figure 4A). The Isc values at steady state less than the average of the currents in the absence of stimulation are quantified in Figure 4C. In all cases, Myo5b kd cells showed significantly less electrogenic response to forskolin. The forskolin response represents the maximum electrogenic activity induced by cAMP. Therefore, it reports on the total available apical anion channels responsive to PKA stimulation. Accordingly, this result indicated a reduction in the number of channels expected from the apical traffic defects resulting from the loss of Myo5b. However, it is important to notice that under Myo5b kd, forskolin-induced currents were reduced only to approximately half of the value in controls (Figure 4C), indicating that a substantial number of apical anion channels remain available for transport in Myo5b-­defective cells.

Because multiple ion transport mechanisms coexist in intestinal epithelial cells, we wanted to simplify the number of electrogenic mechanisms by using a modified Ringer medium, without amino acids and supplemented with glucose on the basolateral side only, to supply energy. To eliminate the apical electrogenic Na⁺/glucose cotransport, glucose was replaced by mannitol on the apical side. Under these conditions, we asked how much of the basal, unstimulated current is carried by Cl⁻. To that end, scrambled or Myo5b shRNA-expressing cells grown with or without dexamethasone, were sequentially measured in normal Cl⁻ Ringer, low Cl⁻ Ringer (gluconate), and back again in normal Ringer. In other words, Cl⁻ concentration was reduced on both sides of the membrane by 100-fold. The basal currents in normal Cl⁻ in all cases were small (3 ± 3.2 µA/cm²). There were no statistically significant differences among all the conditions. Chloride replacement resulted in small changes in basal unstimulated currents, indicating that the current was not carried by Cl⁻. The exceptions were Myo5b kd cells incubated in dexamethasone (Figure 4, B, black trace, and D). In this case, a statistically significant change in transepithelial current was observed, opposite to the current generated by electrogenic apical Cl⁻ secretion (e.g., Figure 4A, forskolin). Similar changes were observed when cells in normal Cl⁻ were supplemented with the PKA inhibitor Rp-cAMP, which blocks the cAMP binding site and is, therefore, highly specific for PKA (Figure 4B). These results support the notion that unstimulated Myo5b-defective intestinal cells display apical anion secretion driving isotonic fluid secretion under PKA signaling in the presence of dexamethasone.

PKA-dependent phosphorylated targets localize to the apical region in Myo5b KO mice and human MVID enterocytes

To confirm whether enhanced PKA activity in the apical domain of intestinal epithelial cells deficient in Myo5b is also observed in vivo, in nontransformed cells, we analyzed a Myo5b KO mouse model and human tissues from MVID patients.

The global Myo5b KO mouse model was previously described (Cartón-García et al., 2015). The pups can suck milk but develop massive diarrhea with hypoglycemia and die within 12 h after birth. Accordingly, samples from these mice were obtained at the late fetal stage when intestinal crypts are not yet developed, and only immature villi are observed. Tissue sections were processed for immunohistochemistry (IHC) using the PKA p-substrate antibody described in Figure 1, C and D. In wild-type (wt) mice, signal appeared in enterocytes near the tip of villi. It was distributed to the cytoplasm and the lateral membrane, and, to some extent, to the apical membrane (Figure 5A, arrow). Conversely, the signal was mostly apical (Figure 5B, arrow) in Myo5b KO samples.

In non-MVID human biopsies, PKA p-substrate signal was diffuse, with a faint apical and lateral membrane concentration (Figure 5D, arrow). On the contrary, in MVID biopsies, the signal was restricted to the apical membrane (Figure 5E, arrow) and to intracellular apical structures consistent with the images of MIs (Figure 5E, arrowhead). In summary, Myo5b KO in a mouse model of MVID, and MVID in human intestine both result in changes in the subcellular distribution PKA-phosphorylated targets. The redistribution of PKA activity in murine and human Myo5b loss of function strongly resembles the observations in Myo5b kd C2BBe (Figure 1D and Supplemental Figure 1B) and T84 cells (Figure 2B). Of note, we conclude that the Myo5b defect results in apical PKA activity with decreased activity in the cytoplasm and lateral membrane and increased activity in the apical domain.

Apical cAMP concentration gradients appear in Myo5b-deficient cells incubated in dexamethasone

If the Myo5b loss-of-function effect on PKA activity has any functional implication in apical CFTR channel function, it must be localized very close to the membrane. Confocal microscopy does not provide adequate resolution in the z-axis to make this assertion. Thus, the images in Figures 1 and 2 could represent PKA targets in subapical compartments, which are enlarged in Myo5b-deficient intestinal cells. Therefore, we wanted to examine whether PKA activity and cAMP localize in the vicinity of the apical CFTR channel (Ameen et al., 2003) in Myo5b loss of function. We confirmed this in the human T84 cells expressing scrambled (scr) and anti-Myo5b shRNA. The latter showed increased colocalization of CFTR and PKA p-substrate signals (Figure 6A, merge). To verify whether this colocalization was within distances enabling molecular interaction, we performed fluorescence resonance energy transfer (FRET) determinations using fluorochromes CY3 and CY5 covalently linked to secondary antibodies. FRET efficiency was measured after acceptor (CY5) photobleaching in random ROIs in the apicalmost confocal optical section (Figure 6B). FRET efficiency was significantly increased 10-fold in Myo5b-deficient cells, preincubated in dexamethasone, as compared with cells expressing scrambled RNA (Figure 6C). While these experiments do not allow calculation of the Förster radius of interaction, we conclude that PKA-phosphorylated sites are concentrated within four immunoglobulin G (IgG) molecule lengths (∼240 nm) of apical CFTR molecules as a result of the Myo5b defect and incubation in dexamethasone.

It is generally accepted that spatially restricted adenylate cyclase activity may result in submicron-sized cAMP microdomains, although the mechanisms to restrict cAMP diffusion remain subject to debate (Calebiro and Maiellaro, 2014).To independently test the hypothesis that PKA activity is increased in the apical domain of Myo5b-deficient cells, we transduced confluent T84 cultures with a cAMP upward fluorescence biosensor with high efficiency using a modified baculovirus that infects mammalian cells. Confluent fully polarized live cells were imaged 48 h after transduction. Low levels of fluorescence, mostly localized to the basolateral region and cytoplasm, were observed in cells expressing scrambled shRNA. However, the fluorescence increased dramatically 30 min following treatment with 10 µM forskolin in the same cells (Figure 7, A and B, scr; examples of the focal plane of ROIs a, apical, b, basal, and cytoplasmic are shown in yellow boxes). Noticeably, cAMP signal was distributed throughout the cytoplasm and not particularly concentrated under the apical domain in forskolin-stimulated cells. The cAMP signal was increased in the apical domain of Myo5b-deficient cells, especially when preincubated in dexamethasone (Figure 7, A and B, kd). It is important to notice that cAMP gradients were inverted from a cytoplasmic/basal larger than apical in nonstimulated cells, to apical larger than basal when Myo5b deficiency was concurrent with dexamethasone supplementation (Figure 7B, arrows). Also, this increase in apical cAMP was observed in the absence of exogenous stimulation other than dexamethasone (serum-free medium and no forskolin). This is consistent with the increase in apical PKA p-substrate signal described in the same region of the cell in two intestinal Myo5b kd cell lines, Myo5b KO mice, and human MVID enterocytes as shown in previous sections.

Myo5b defect along with dexamethasone increase PKA-dependent phosphorylation in apical membrane proteins in intestinal epithelial cells

To confirm that apical membrane proteins, which would include CFTR, are phosphorylated in PKA targeted sites, cells were screened by vectorial cell surface biotinylation (Rodriguez-Boulan et al., 1989). Not all proteins can be biotinylated. Derivatization depends on the presence of primary amines in the amino acid side chains exposed to the extracellular milieu. Therefore, surface biotinylation provides a sampling of apical surface proteins that can be derivatized with sulfo-NHS-biotin. As expected from deficient apical membrane traffic, the surface expression of a subset of apical proteins decreased in C2BBe Myo5b-deficient cells. Some bands were undetectable in Myo5b kd cells, while others were only very moderately decreased (Figure 8A, stars). This result is consistent with previous observations that loss of Myo5b leads to selective apical traffic defects differentially affecting individual apical membrane protein surface expression (Kravtsov et al., 2016).

To extend this result to the other cell line and to understand the effect of Myo5b defect and GCs on apical membrane proteins, similar experiments were conducted. T84 cell cultures were incubated in dexamethasone (or vehicle) for 10 d. When Triton X-100 extracts of biotinylated (or not, control c) cells were analyzed, the decrease in apically exposed proteins in Myo5b kd was confirmed, but little or no effect of GCs was noticed (Figure 8B). To determine whether PKA targeting of biotinylated apical membrane proteins is affected by loss of Myo5b or corticoids, pull downs with streptavidin Sepharose beads followed by PKA p-substrate immunoblot were performed. Pull down resulted in enrichment in fewer bands phosphorylated by PKA, not readily noticeable in the total extract. No major bands were observed in the 140–200 kDa range, so we focused the analysis on two major bands with M_r 54 and 60 × 10³ (Figure 8C, arrows) in the absence of GC supplementation; phosphorylation remained at similar levels in Myo5b-deficient cells. However, for cells grown in the presence of GC, Myo5b loss of function resulted in a statistically significant increase in phosphorylation (Figure 8, C and D). These results were quantified by densitometry of the two major bands (Figure 8D). Considering that expression of apical proteins decreases on the surface to various extents (Figure 8, B and C) because of membrane traffic defects dependent on Myo5b loss of function, these results indicate that proteins reaching the apical surface are exposed to a heightened PKA kinase activity in cells deficient in Myo5b and exposed to physiological levels of GCs.

The previous result raises the possibility that CFTR may be a target of PKA in the apical surface in Myo5b-deficient cells. To test that hypothesis, we took advantage of the previous work of various groups showing that 14-3-3 proteins associate with active CFTR through multiple PKA target phosphosites in tandem in the intrinsically disordered regulatory domain (Bozoky et al., 2013) stabilizing the protein (Stevers et al., 2016). Importantly, CFTR coimmunoprecipitates with 14-3-3, and co-ip is enhanced by forskolin (Liang et al., 2012). In other words, 14-3-3β and CFTR coimmunoprecipitation is a reporter of PKA phosphorylation. In enterocytes and intestinal cell lines, 14-3-3β is expressed (Monroy, 2008; Gomez-­Suarez et al., 2016), along with isoforms ε, θ, ζ, and η. In addition, it has been shown that 14-3-3β heterodimerizes with ζ (Alvarez et al., 2003) and ε (Ghorbani et al., 2016). Because we did not determine the specific isoforms (homo- or heterodimers) involved in this coimmunoprecipitation, we will generically refer to 14-3-3 proteins hereafter. These proteins are abundant and cytosolic. The fraction bound to phosphorylated membrane proteins, thus, is expected to be relatively small. Accordingly, we enriched our extracts in membrane proteins by using Triton X-114 extraction, instead of the standard Triton X-100, followed by temperature separation of the detergent phase (Bordier, 1981; Taguchi and Schatzl, 2014). As predicted, Myo5b kd resulted in a steep increase in CFTR coimmunoprecipitation with 14-3-3 (Figure 8E). Interestingly, when this co-ip is driven by forskolin, both immature (b) and mature (glycosylated, c) forms of CFTR coimmunoprecipitate with 14-3-3 (Liang et al., 2012). In the case of Myo5b kd cells, however, it was mostly the c form of CFTR that coimmunoprecipitated with 14-3-3, suggesting that phosphorylation occurred in a compartment beyond the Golgi apparatus. This result supports the notion that the fraction of CFTR that reaches the surface in Myo5b-defective epithelial cells is heavily phosphorylated in the regulatory domain in Myo5b-deficient cells, especially in the presence of GCs.

Antibodies and reagents

Anti-PKA–phosphorylated substrate sites (PKA p-substrate, anti- RRXpS/pT) antibody was obtained from Cell Signaling (cat. # 9621S). The sources of other antibodies were as follows: GAPDH (Calbiochem; CB1001); tubulin (Abcam; ab52866); Myo5b (Novus Biologicals; NBP1-87746); ZO-1 MAb (ZO-1-1A12; ThermoFisher Scientific). CFTR monoclonal antibodies 596 and 769, against the NBD2 (C-terminal) domain (Kreda et al., 2005; Cui et al., 2007) were obtained from the Cystic Fibrosis Foundation, CFTR Antibody Distribution Program, University of North Carolina at Chapel Hill. Anti-14-3-3 (Proteintech; AB_2218096) was originally developed against full-length 14-3-3θ, but it cross-reacts with other 14-3-3 isoforms (Bai et al., 2014) and has been validated for immunoprecipitation. Fluorescent secondary affinity-purified antibodies with minimal cross-reactivity were obtained from Jackson Immunoresearch Laboratories. ECL horseradish peroxidase (HRP)–conjugated affinity-purified antibodies were obtained from Amersham. TrueBlot HRP–anti-rabbit IgG antibody recognizing native conformation nonreduced protein was purchased from Rockland (TruBlot; 18-8816-31).

Cells and lentivirus

CaCo2 BBe(C2BBe) and T84 cells were obtained from the American Type Culture Collection. Lentivirus particles expressing anti-Myo5b shRNA or GFP were described before (Kravtsov et al., 2014). Upon transduction, both cell types were continuously selected in 20 µg/ml puromycin. For experiments, the cells were kept in 12.5 µg/ml puromycin. Cells were kept for up to 10 passages after transduction or until the efficiency of knockdown was less than 90%. For experiments, C2BBe or T84 cells were grown on 3-µm-pore Transwell or Snapwell filters (Costar) for 2 wk after confluency, when C2BBe cells reach full differentiation. In preliminary experiments, we confirmed by reverse transcription-quantitative PCR that T84 cells also acquire the highest CFTR expression at that time.

T84 three-dimensional cultures (spheroids) in Matrigel (BD Biosciences) were prepared following the technique described elsewhere for Madin–Darby canine kidney cells (Martin-Belmonte et al., 2008), and adapted for T84 cells (Bentz et al., 2013). The spheroids were used for experiments after 10 d in culture, with changes in media every other day.

Dexamethasone and pharmacological treatments

Synthetic glucocorticoid dexamethasone (Sigma) was kept in dimethyl sulfoxide (DMSO) stock solution for up to 3 mo at −20°C. Dexamethasone concentration in treatments was established on the basis of average blood cortisol concentrations in neonates, which range from 40 µg/l (∼0.1 µM; Jonetz-Mentzel and Wiedemann, 1993) to 250 µg/l (∼0.5 µM; Stevens, 1970), depending on the assays. Typically, cells were incubated in dexamethasone for 10 d before the day of the experiment to mimic the timing of the increase in serum GCs before birth.

Forskolin, Rp-cyclic 3′,5′-hydrogen phosphorothioate adenosine (Rp-cAMPS), and CFTR172 were purchased from Sigma. All these reagents, except Rp-cAMP, were kept in DMSO stock solutions at −20°C. Rp-cAMP stock solution was made in sterile phosphate-buffered saline (PBS) and frozen. H89 was purchased from Tocris.

Human samples were considered exempt nonhuman research on the basis that they were deidentified paraffin blocks retrieved from a repository.

Mouse intestine

The generation of the Myo5b knockout mouse model used has been described before (Cartón-García et al., 2015). Mice heterozygous for the Myo5b-null allele were crossed and pregnant females were killed by CO₂ asphyxiation at day 20 of gestation. The embryos were obtained, kept on ice-cold PBS, and killed by decapitation. Tail clips were taken for DNA extraction and genotyping and the whole embryos were Formalin-fixed and paraffin-embedded as described (Cartón-García et al., 2015).

IHC and immunofluorescence for PKA p-substrate antibody

Immunohistochemistry for the PKA p-substrate antibody was performed using the antigen retrieval protocol described by the manufacturer (www.cellsignal.com/products/primary-antibodies/phospho-ser-thr-pka-substrate-antibody/9621). Published data include validation by negative controls using phosphatases before tissue staining (same website). Independent validation of the use of this antibody for immunohistochemistry was published by others (Saitoh et al., 2009). For immunofluorescence, the antibody was used after fixation in 10% trichloroacetic acid as described for other antibodies against phosphorylated epitopes (Hayashi et al., 1999; Wald et al., 2011).

Vectorial apical cell surface biotinylation was performed as described before (Rodriguez-Boulan et al., 1989; Sargiacomo et al., 1989). Briefly, confluent cell monolayers grown on Transwell filters were washed in PBS from the apical side only, leaving normal culture medium in the basal chamber. The cells were incubated on ice with 1 mM sulfo-NHS-LC-biotin (ThermoFisher Scientific; A39256) for 15 min, and then extensively washed with ice-cold DMEM. The cells were extracted in RIPA buffer in the presence of anti-proteases and anti-phosphatases. After centrifugation, the supernatants were incubated in streptavin-agarose beads (Sigma Chemical; S1638) for pull down. It must be noted that RIPA buffer extraction, while efficient for most membrane proteins has shown poor yields specifically for CFTR.

Immunoprecipitation of CFTR 14-3-3 complexes was performed as described before (Kravtsov et al., 2012) but replacing Triton X-100 with 2% Triton X-114 to enrich for transmembrane 14-3-3 interactors, as described elsewhere (Bordier, 1981). It must be highlighted that CFTR is poorly extracted in RIPA buffer. An ad hoc CFTR extraction buffer (25 mM HEPES, 10% glycerol, pH 7.4, supplemented with protease inhibitors, HG) was used for these experiments as described before (Collaco et al., 2010). In this buffer Triton X-114 clouding point is around 30°C. The 6% Triton X-114 HG stock solution was subjected to at least five cycles of condensation discarding the aqueous phase and replaced by freshly prepared HG buffer each time (Taguchi and Schatzl, 2014). This precondensed stock was diluted 1:3 in freshly prepared HG buffer before cell extraction (EB). Cell monolayers were washed with PBS on ice and extracted in EB for 15 min. The extracts were spun at 12,000 × g in the cold for 15 min and the supernatant was warmed to 33°C to separate the Triton X-114 phase (1 h). The aqueous supernatant was discarded, and the lower detergent phase resuspended in the same volume of cold HG buffer. Samples of this extract were taken as input. Incubation with the anti-14-3-3 antibody was performed in EB overnight at 4°C. Sepharose–protein A beads were prewashed in the same buffer for 1 h, spun, and added to the cell extracts for 3 h. Upon centrifugation, the beads were further washed in 1% Triton X-100 in PBS four times (1 h), and finally in 0.5M Tris buffer, pH 6.8, for 10 min. The protein was eluted in SDS sample buffer supplemented with 0.1% 2-mercaptoethanol at 37°C for 10 min. After separating the beads, the procedure was repeated one more time and the eluates were pooled.

cAMP biosensor

A cAMP upward mNeon Green cADDis biosensor (Tewson et al., 2016) expressed under a CMV promoter was obtained from Montana Molecular (U0200G) and delivered by transduction with BacMam, a baculovirus modified to infect mammalian cells (Kost et al., 2007). Fully differentiated T84 cell cultures (2 wk after confluency) grown on Transwell filters were transiently transduced with high efficiency by this vector. The cells were incubated in the presence of 2 mM sodium butyrate for 48 h after transduction according to the manufacturer’s specifications before imaging. Six hours before imaging, 0.5 µg/ml 4′,6-diamino-2-phenylindole was added to the medium to contrast nuclei. Some cultures were supplemented with 10 µM forskolin for 30 min as positive control. Live cells were imaged in an inverted confocal microscope using a glass-bottom Petri dish. During imaging the cells were kept at 37°C and incubated in serum-free phenol red–free DMEM buffered with 20 mM HEPES.

Confocal microscopy

Confocal microscopy images were obtained in a Leica SP5 using an oil immersion 63× objective. For live-cell microscopy a water immersion 63× objective was used. Unless stated otherwise, the sections were collected at Airy = 0.9 or Airy = 1.2 for live cells to decrease laser intensity. For Airy 0.9, the section thickness was typically 0.6 µm and section spacing was 0.5 µm. For three-dimensional reconstructions, the stacks were imported into Slidebook software and cropped into 10-pixel-wide images. These images were reconstructed and rotated to show the z-axis. For quantification, ROIs were identified at random by overlapping a grid on the image in original (XY) images without rotation. To ensure that the measurements were performed at the apical domain, all the ROIs used for quantification comprised images of microvilli.

Live-cell microscopy was performed within a warmed chamber at 5% CO₂ using a Leica DMI6000 inverted microscope and a 20× objective. Time-lapse acquisition was done with LAS X software with images acquired every 15 min.

Fluorescence resonance energy transfer

FRET indexes were measured for the pair CY3-CY5 using the acceptor bleaching technique (Berney and Danuser, 2003; Centonze et al., 2003). The method followed an established protocol for FRET using labeled antibodies (Snapp and Hegde, 2006). The efficiency of energy transfer was calculated as E = 1−I_DA/I_D, where I_DA is the average pixel intensity of the donor in the presence of the acceptor (before photobleaching) and I_D is the fluorescence intensity of the same pixel in the absence of donor (after full photobleaching). FRET determinations were performed on fixed and permeabilized cells incubated with antibodies against CFTR (mouse monoclonal) and phospho-PKA substrate (rabbit monoclonal) followed by affinity-purified antibodies labeled with Cy3 and Cy5. Confocal images were acquired as described above at Airy = 1, which, for the 63× 1.3 NA objective, represented 0.7-µm-thick confocal sections. ROIs were randomly distributed over the last confocal plane where an apical image was distinguishable in XY sections. Better than 90% acceptor photobleaching was reached after 200 iterations of the 633-nm laser beam scan when the cells were mounted in medium without anti-fade reagents. Anti-fade 1% n-propyl gallate was used in the mounting medium in all other immunofluorescence procedures.

Transepithelial fluid transport

Net fluid fluxes were measured gravimetrically as described before (Kravtsov et al., 2014). Briefly, cell confluent, differentiated cell monolayers were grown on 3-µm-pore Transwell filters. Fluid tightness was determined by filling up the apical chamber with tissue culture medium, which should keep the volume for 24 h before the flux measurements. Apical and basal chambers were then equilibrated at the same level to minimize hydrostatic pressure. At this time, drugs were added to the apical chamber. Upon careful drying of the Transwell outer surfaces, the filters were weighted with 0.5 mg precision, which is <5% error for most measurements. Measurements in triplicate were repeated at various times. In the presence of standard concentrations of antibiotic–antimycotic (Life Technologies), no visible contamination was present after 24 h when the experiment was concluded. The order of weighting filters for different experimental conditions was randomized in different independent experiments.

For 3D cultures, dexamethasone or DMSO vehicle was added to the medium outside the Matrigel culture and the image acquisition was initiated 3 h later. The maximum caliper diameter at time 0 and at 23 h was taken for each spheroid, and the difference recorded as percent change.

Transepithelial short circuit currents (Isc)

T84 cells, transduced with lentivirus expressing scrambled or anti-Myo5b shRNAs, were grown on 0.4-µm-pore Snapwell filters to confluency and supplemented or not with dexamethasone for 14 d. Puromycin was omitted from culture media when the cells became confluent. Snapwells were inserted in Ussing chambers with continuous air bubbling containing either complete DMEM culture medium or modified Ringer solutions. Temperature was kept at 37°C by an external metal jacket. Transepithelial current and voltage were recorded for an initial measurement of resistance, and then potential was clamped. Measurements were performed using two separate pairs of Ag/AgCl electrodes via 2% agar bridges containing 3M KCl. Recordings of transepithelial resistance and short circuit currents were performed with a Physiologic Instruments VCC MC6 meter via a Biopac Systems digitizer using Acknowledge software. PKA agonist (10 µM forskolin) and antagonist (40 µM Rp-cAMP) were added to the apical side only. For experiments determining Isc in the presence of PKA inhibitors, the following solutions were used: Modified Ringer (MR) 10 mM HEPES, 5 mM NaH₂PO₄, 0.12 mM CaCl₂, 1.2 mM MgCl₂; normal Cl⁻ Ringer was MR supplemented with 128 mM NaCl; and low Cl⁻ Ringer was MR supplemented with 128 mM sodium gluconate and 0.1 mM CaCl₂. In all cases, pH was 7.4. All Ringer solutions applied to the apical chamber (either normal or low Cl⁻) were supplemented with 10 mM mannitol, while solutions applied to the basolateral (“serosal”) chamber were supplemented with 10 mM glucose.

Statistics

Significance was estimated by t test for normally distributed variables or by one-way analysis of variance (ANOVA) for significance of multiple treatments. Otherwise, nonparametric Kruskal-Wallis test was used.