RNA Interference Screen to Identify Kinases That Suppress Rescue of ΔF508-CFTR

Abstract

Cystic Fibrosis (CF) is an autosomal recessive disorder caused by mutations in the gene encoding the Cystic fibrosis transmembrane conductance regulator (CFTR). ΔF508-CFTR, the most common disease-causing CF mutant, exhibits folding and trafficking defects and is retained in the endoplasmic reticulum, where it is targeted for proteasomal degradation. To identify signaling pathways involved in ΔF508-CFTR rescue, we screened a library of endoribonuclease-prepared short interfering RNAs (esiRNAs) that target ∼750 different kinases and associated signaling proteins. We identified 20 novel suppressors of ΔF508-CFTR maturation, including the FGFR1. These were subsequently validated by measuring channel activity by the YFP halide-sensitive assay following shRNA-mediated knockdown, immunoblotting for the mature (band C) ΔF508-CFTR and measuring the amount of surface ΔF508-CFTR by ELISA. The role of FGFR signaling on ΔF508-CFTR trafficking was further elucidated by knocking down FGFRs and their downstream signaling proteins: Erk1/2, Akt, PLCγ-1, and FRS2. Interestingly, inhibition of FGFR1 with SU5402 administered to intestinal organoids (mini-guts) generated from the ileum of ΔF508-CFTR homozygous mice resulted in a robust ΔF508-CFTR rescue. Moreover, combination of SU5402 and VX-809 treatments in cells led to an additive enhancement of ΔF508-CFTR rescue, suggesting these compounds operate by different mechanisms. Chaperone array analysis on human bronchial epithelial cells harvested from ΔF508/ΔF508-CFTR transplant patients treated with SU5402 identified altered expression of several chaperones, an effect validated by their overexpression or knockdown experiments. We propose that FGFR signaling regulates specific chaperones that control ΔF508-CFTR maturation, and suggest that FGFRs may serve as important targets for therapeutic intervention for the treatment of CF.

Cystic fibrosis (CF)¹ is a pleiotropic disease caused by an abnormal ion transport in the secretory epithelia lining the tubular organs of the body such as lungs, intestines, pancreas, liver, and male reproductive tract. In the airways of CF patients, reduced Cl⁻ and bicarbonate secretion caused by lack of functional Cystic fibrosis transmembrane conductance regulator (CFTR) on the apical surface, and hyper-absorption of Na⁺ because of elevated activity of ENaC (1), lead to a dehydration of the airway surface liquid (ASL). This reduces the viscosity of the mucus layer and the deposited layer of thickened mucus creates an environment that promotes bacterial colonization, which eventually leads to chronic infection of the lungs and death (2, 3).

CFTR is a transmembrane protein that functions as a cAMP-regulated, ATP-dependent Cl⁻ channel that also allows passage of bicarbonate through its pore (4, 5). It also possesses ATPase activity important for Cl⁻ conductance (6, 7). The CFTR structure is predicted to consist of five domains: two membrane spanning domains (MSD1, MSD2), each composed of six putative transmembrane helices, two nucleotide binding domains (NBD1, NBD2), and a unique regulatory (R) region (8).

More than 1900 CFTR mutations have been identified to date (www.genet.sickkids.on.ca/cftr). The most common mutation is a deletion of phenylalanine at position 508 (ΔF508 or ΔF508-CFTR) in NBD1 (9). The ΔF508 mutation causes severe defects in the processing and function of CFTR. The protein exhibits impaired trafficking from the endoplasmic reticulum (ER) to the plasma membrane (PM), impaired intramolecular interactions between NBD1 and the transmembrane domain, and cell surface instability (10–15). Nevertheless, the ΔF508 defect can be corrected, because treating cells expressing ΔF508-CFTR with low temperature or chemical chaperones (e.g. glycerol) can restore some surface expression of the mutant (11, 16).

Numerous small molecules that can at least partially correct (or potentiate) the ΔF508-CFTR defect have been identified to date (17–27), and some were already tested in clinical trials (e.g. sildenafil, VX-809/Lumacaftor), or have made it to the clinic (VX-770/Kalydeco/Ivacaftor) (http://www.cff.org/research/DrugDevelopmentPipeline/). However, the need to identify new ΔF508-CFTR correctors remains immense as the most promising corrector, VX-809, has proven ineffective in alleviating lung disease of CF patients when administered alone (27). Thus, our group developed a high-content technology aimed at identifying proteins and small molecules that correct the trafficking and functional defects of ΔF508-CFTR (28). We successfully used this approach to carry out three separate high-content screens: a protein overexpression screen (28), a small-molecule kinase inhibitor screen (29) and a kinome RNA interference (RNAi) screen, described here.

EXPERIMENTAL PROCEDURES

Media and Reagents

Dulbecco's Modified Eagle's Medium (DMEM), Dulbecco's Phosphate Buffered Saline (d-PBS), Fetal Bovine Serum (FBS), trypsin, G418, blasticidin, and zeocin were obtained from Invitrogen (Carlsbad, CA). The mouse M3A7 anti CFTR monoclonal antibody was purchased from Millipore, Temecula, CA, the mouse HA.11 (16B12) monoclonal antibody was from Covance (San Diego, CA)., the rabbit polyclonal antivinculin antibody was from Abcam (Cambridge, UK), and SuperSignal West Femto Maximum Sensitivity kit was from Pierce (Rockford, IL). The High Capacity cDNA Reverse Transcription kit was obtained from Applied Biosystems (Foster City, CA), the Platinum® SYBR® Green qPCRSuperMix-UDG was from Invitrogen, and the SA-HRP was from eBioscience. The kinome RNA interference (esiRNA) library was obtained from Dr. Laurence Pelletier (The Samuel Lunenfeld Research Institute - see below). shRNA clones were from the RNAi Consortium (TRC) (30) and SIDNET/SPARC BioCentre, The Hospital for Sick Children (pGIPZ). For the overexpressed chaperones, the entry clones compatible with Gateway® system were obtained from SIDNET/SPARC BioCentre and PlasmID (The Dana-Farber/Harvard Cancer Center DNA Resource Core), and were subsequently cloned into the destination vector, pcDNA3.1(eYFP H148Q/I152L). All constructs were sequence verified for accurate coding.

Cloning and Cell Line Generation

The triple hemagglutinin (3HA) tag was cloned by PCR based on the sequence used in BHK ΔF508-CFTR 3HA cell line (22), and inserted after Asn at position 901 (N901) in the fourth external loop of CFTR. The full length CFTR bearing 3HA tag (wild type or ΔF508) was subsequently cloned into the pLVE/zeo vector. The plasmids were then transfected into the HEK293 MSR GripTite using calcium phosphate precipitation. The transfected cells were seeded at different concentrations to isolate individual colonies under selection with 100 μg/ml zeocin. Individual clones were picked, expanded and CFTR or ΔF508-CFTR expression verified by immunoblotting (supplemental Fig. S1).

Cellomics YFP Halide-exchange Assay

Kinome esiRNA Library

The kinome cassette of a genome-scale library of endoribonuclease-prepared short interfering (esi)RNAs (30) contains dsRNA pools that target 759 different kinases and associated signaling proteins (supplemental Table S1).

esiRNA Screen and Drug Combination Testing Using the Cellomics YFP Halide-Exchange Assay

HEK293 MSR GripTite (293MSR-GT) cells stably coexpressing eYFP(H148Q/I152L) and ΔF508-CFTR proteins (29) were cultured in DMEM medium supplemented with 10% FBS, 1× Nonessential Amino Acids, 0.6 mg/ml G418, 10 μg/ml blasticidin, and 50 μg/ml zeocin, at 37 °C, 5% CO₂ in humidified atmosphere.

The Cellomics halide-exchange assay was performed as described previously (29). Briefly, 5 × 10⁴ ΔF508-CFTR cells (i.e. 293MSR-GT cells stably coexpressing eYFP(H148Q/I152L) and ΔF508-CFTR) per well were seeded in 96-well plates. The next day the cells were transfected with esiRNA duplexes from the library (final concentration 40 nm), luciferase (nonsilencing control), EG5 (transfection control), or AHA1 esiRNA (assay control), using Lipofectamine 2000. Medium was changed 6 h after transfection, and the cells were placed at 37 °C, 5% CO₂ for 72 h. The 96-well transfection protocol was optimized using EG5 (KIF11) esiRNA as a transfection control. The transfection was considered successful if more than 80% of ΔF508-CFTR cells exhibited round-shape phenotype 72 h post-transfection (see “Results”). After 72 h of incubation, the medium was replaced with 152 μl of chloride solution (137 mm NaCl, 2.7 mm KCl, 0.7 mm CaCl₂, 1.1 mm MgCl₂, 1.5 mm KH₂PO₄, 8.1 mm Na₂HPO₄, pH 7.1), in the absence or presence of FIG (25 μm Forskolin, 45 μm IBMX, 50 μm Genistein), at 37 °C. After 20-min incubation, 92 μl of iodide buffer (137 mm NaI, 2.7 mm KCl, 0.7 mm CaCl₂, 1.1 mm MgCl₂, 1.5 mm KH₂PO₄, 8.1 mm Na₂HPO₄, pH 7.1) was added (final I⁻ concentration 52 mm). Using the Cellomics KineticScan KSR Reader (Thermo Fisher) and a modified Target Activation algorithm, objects (individual cells or sometimes clusters of cells) were defined by eYFP(H148Q/I152L) fluorescence intensity, and the fluorescence quenching over 24-s time course at 30 °C, 5% CO₂, was recorded. Valid wells contained between 70 and 300 objects per field (single field per well). Genes that displayed a difference in the YFP fluorescence intensity (between FIG-stimulated sample and nonsilencing control) lower than 0.09 were rejected after the first two runs of the screen. This cut-off value equaled three times the standard deviation from the mean value of the control (AHA1). The rest of the esiRNA duplexes (56 genes) were subjected to the third run of the screen. Twenty top hits of the screen were subjected to further validation of ΔF508-CFTR rescue by functional assay (Cellomics ArrayScan VTI platform) immunoblotting and ELISA following shRNA-mediated knockdown.

For drug combination testing, 8 × 10⁴ ΔF508-CFTR cells (i.e. 293MSR-GT cells stably coexpressing eYFP(H148Q/I152L) and ΔF508-CFTR) per well were seeded in 96-wells plates. The next day, the cells were treated with either SU5402+VX809 or AZD4547+VX809 with concentration ranging from 1 μm to 10 μm. The cells were then incubated at 37 °C, 5% CO₂ for 48 h, and analyzed by the Cellomics halide-exchange assay, as above.

Validation of the esiRNA Screen

Cellomics Analysis Following shRNA Knockdown

Prior to the Cellomics halide-exchange assay ΔF508-CFTR cells (stably expressing eYFP(H148Q/I152L)) were transfected with shRNA constructs targeting the identified genes or luciferase (nonsilencing control), using Lipofectamine 2000, according to the manufacturer's instructions. Medium was changed 6 h after transfection and ΔF508-CFTR cells were placed at 37 °C, 5% CO₂. Forty-eight hours after transfection the cells were incubated with media containing puromycin (5 μg/ml, 3 days). Cellomics halide-exchange assay was performed as described above, using Cellomics ArrayScan VTI platform (data from three fields per well). A total number of 133 shRNA clones was screened (multiple shRNA clones per gene) (supplemental Table S2). Knockdown efficiency was validated by two-step RT-qPCR as described previously (29). Briefly, total RNA was isolated using the RNeasy 96 kit (Qiagen, Venlo, Netherlands), and cDNA was prepared using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems). Real-time PCR reactions were performed using Platinum® SYBR® Green qPCRSuperMix-UDG (Invitrogen) and CFX96 Real-Time System (BioRad). Primers were obtained from Integrated DNA Technologies. For standard curves, real-time PCR was performed on a fivefold dilution series of pooled cDNA.

Immunoblotting

The appearance of band C was validated by immunoblotting, as described previously (28, 29). Briefly, prior to immunoblotting ΔF508-CFTR cells were transfected with shRNA constructs (TRC) for the identified genes or nonsilencing (scrambled) control for 48 h at 37 °C, or incubated for 48 h at 27 °C (positive control). After transfection, the cells were incubated with media containing puromycin (5 μg/ml, 3 days). ΔF508-CFTR cells were then rinsed in cold PBS and lysed in lysis buffer (50 mm Hepes, pH7.5, 150 mm NaCl, 1.5 mm MgCl₂, 1 mm EGTA, 10% glycerol (v/v), 1% Triton X-100 (v/v), 2 mm PMSF, 2× PAL inhibitors). Proteins were resolved on SDS-PAGE, transferred to nitrocellulose membranes and immunoblotted with anti CFTR (M3A7, 1:1000) or anti vinculin (1:2000) antibodies. Membranes were washed with 5% Blotto, incubated with HRP-conjugated anti-mouse or anti-rabbit antibodies (1:10,000) and washed with PBST. Signal was detected with SuperSignal West Femto reagent.

ELISA

ΔF508-CFTR 3HA cells were transfected with shRNA constructs (pGIPZ) for the identified genes or nonsilencing control, using Lipofectamine 2000, according to the manufacturer's instructions. Forty-eight hours after transfection the cells were incubated with media containing puromycin (5 μg/ml, 3 days). The cells were then biotinylated with 0.5 mg/ml biotin in PBS (15 min), washed with ice-cold PBS and lysed. To capture CFTR or ΔF508-CFTR, 50 μg of total lysate protein (per well) were incubated with anti HA antibody (1:400) in 96-well plate, for 2 h at 4 °C. The plates were then washed with PBST (PBS + 0.05% Tween) and SA-HRP (1:1000) was added in ELISA buffer (PBST + 0.5% BSA) into each well (20 min). After washing, the plates were developed with TMB substrate. The reaction was stopped with 1N H₂SO₄ and the plates were read at 450 nm.

Short-circuit Current Analysis in MDCK Cells Stably Expressing ΔF508-CFTR Mutant

MDCK cells stably expressing ΔF508-CFTR were infected with pGIPZ lentiviral particles for the 10 human-to-canine compatible shRNAs (CDK10, PANK1, PANK4, RPS6KC1, DUSP22, SOCS1, FGFR1, CLK3, NEK10, BRAF, PCK2, IPMK), using polybrene (SIDNET Lentiviral Facility, The Hospital for Sick Children). The cells were subjected to puromycin selection (5 μg/ml), and 5 × 10⁵ infected cells were seeded onto 1.12 cm² snapwells (Corning). After 6–7 days the inserts were mounted in Ussing chambers (Physiological Instruments, San Diego, CA) and studied under voltage clamp conditions as previously described (29). Briefly, ENaC channels were inhibited with 10 μm amiloride (Sigma, St. Louis, MO) and non-CFTR chloride transporters were blocked with 250 μm DNDS (4,4-dinitrostilbene-2,2-disulfonate, Sigma). CFTR currents were then stimulated using 25 μm Forskolin, 25 μm IBMX and 50 μm Genistein (FIG), and after the indicated time (min) inhibited using 15 μm GlyH-101 (Gly). Data were recorded and analyzed using Analyzer 2.1.3. Knockdown efficiency was validated by two-step RT-qPCR, as described above.

Salivary Secretion Assay (SSA)

The salivary secretion assay, described previously (31), was modified as follows. Male ΔF508 mice (CFTR^tm1Eur on a 129/FVB background) and their wild-type littermates (kindly provided by Dr. C. Bear) of 9–12 weeks were intra-peritoneally injected with DMSO or SU5402 (dissolved in DMSO at the concentration of 6 mg/ml) at 25 mg/kg body weight, every day for 1 week. The mice were weighed daily and the dosages adjusted accordingly. The mice were then anesthetized by inhaling isoflurane until the end of the procedure. Cholinergic antagonist, atropine (1 mm, 50 μl) was subcutaneously injected into the right cheek to block potential cholinergic stimulation of the salivary gland. A small strip of filter paper was placed against the injected cheek, for 4 min. Isoprenaline (10 mm, 37.5 μl) was subsequently injected in the same spot to stimulate an adrenergic secretion of saliva (time 0). Filter strips (preweighed in an Eppendorf tube) were replaced every 5 min, over a period of 30 min. All six filter strips were weighed at the end of the collection and the results were normalized relative to mg/g body weight. All animal work was done in accordance with SickKids Institutional guidelines and approval of the Animal Care Committee.

Intestinal Organoids Experiments

Intestinal organoids derived from crypts isolated from the terminal ileum of ΔF508/ΔF508-CFTR mice and wild-type littermates were generated and maintained in culture, as described (32). For forskolin-induced swelling (FIS) experiments, organoids were seeded in 24-well tissue culture plates, pretreated with kinase inhibitors (SU5402, 10 μm) and/or VX-809 (3 μm), and stimulated with 5 μm forskolin, as outlined in (33). FIS was observed by brightfield live-cell microscopy with an automated xy-stage (Nikon TE-2000 with Solent Scientific enclosure, 20×).

Chaperone Array Screen

Human Bronchial Epithelial (HBE) cells from ΔF508/ΔF508-CFTR transplant patients (P2 cells) were obtained from the University of Iowa Cell Culture Facility and grown on collagen-coated permeable millicell inserts. The cells were treated with DMSO (control), 1 μm or 10 μm SU5402 for 48 h prior to RNA extraction. Total RNA was extracted using the PureLink RNA Mini Kit (Invitrogen) and cDNA was synthesized from 1 μg of mRNA using the High capacity cDNA reverse transcription kit (Applied Bioscience) according to the manufacturer's instructions. Array analysis was performed using the RT² Profiler™ PCR Array Human Heat Shock Proteins & Chaperones kit (Qiagen). mRNA expression levels were determined relative to actin, GAPDH and B2M using the ΔC_t method. Changes in chaperone expression level relative to DMSO control were determined using the ΔΔC_t method. The chaperone array experiment was performed three times and average values are shown in a heat map.

Validation of the Chaperone Array Hits

7 × 10⁵ ΔF508-CFTR 3HA cells per well were seeded in a 6-well plate format. The next day the cells were transfected with the clones for the analyzed chaperone genes (shRNA or overexpression) or luciferase control, using PolyJet™ DNA In Vitro Transfection Reagent according to the manufacturer's instructions. Forty-eight hours post-transfection, the cells that were transfected with shRNA were further incubated with media containing puromycin (5 μg/ml, 3 days). The cells that were transfected with the chaperone overexpression clones were biotinylated, and ELISA was performed as described above.

HSF1 Experiments

The pcDNA3.1(eYFP H148Q/I152L) plasmid containing the wild-type HSF1 was used to construct a constitutively active mutant of HSF1 (34) using site directed mutagenesis consisting of one-step PCR with two overlapping internal primers at the mutagenic site. The internal primers used were 5′GAACGACAGTGGCTCAGCACATGGGCGCCCATCTTCCGTGGAC3′ and 5′GTCCACGGAAGATGGGCGCCCATGTGCTGAGCCACTGTCGTTC3′. DNA sequencing was performed to verify the constructs. 7 × 10⁵ ΔF508-CFTR 3HA cells per well were seeded in 6-well plates. The next day the cells were transfected with the constructs for wild-type HSF1, mutant HSF1, or luciferase control using PolyJet™ DNA In Vitro Transfection Reagent, according to the manufacturer's instructions. Forty-eight hours after transfection, the cells were biotinylated, and ELISA was performed as described above.

RESULTS

Identification of Kinases and Associated Signaling Proteins That Suppress Rescue of ΔF508-CFTR

Delineation of pathways and proteins that prevent rescue of ΔF508-CFTR is important for the identification of drugs that target these pathways. Our group previously developed a high-content functional screen to identify ΔF508-CFTR correctors (and potentiators) in multiple individual cells simultaneously, using Cellomics KineticScan KSR and ArrayScan VTI platforms (28, 29). The kinome esiRNA screen presented in this study complements the small-molecule kinase inhibitor screen described previously (29).

We screened a library of esiRNA duplexes targeting 759 different kinases and associated proteins (supplemental Table S1). Because esiRNA transfection efficiency is critical to performing this screen, and the esiRNA molecules do not carry a selection marker, the 96-well transfection protocol was optimized using EG5 (KIF11) esiRNA as a transfection control added to parallel wells on the same plate. Knockdown of EG5, a kinesin-related motor protein, leads to mitotic arrest and rounded-up phenotype of the cells. The transfection was considered successful if more than 80% of ΔF508-CFTR cells exhibited round-shape phenotype 72 h post-transfection.

In the Cellomics halide-exchange assay the fluorescence quenching of YFP, corresponding to Cl⁻/I⁻ exchange via ΔF508-CFTR, was quantified in cells stably expressing ΔF508-CFTR and transfected with esiRNA duplexes from the library. Fig. 1 depicts several representative “hit” suppressors that when knocked-down exhibited various degrees of correction of the ΔF508-CFTR defect. The complete list of the hits, defined as those exhibiting a difference in average fluorescence intensity ΔFI_avg (between Forskolin-IBMX-Genistein (FIG)-stimulated sample and FIG-stimulated nonsilencing control) of at least 9%, is provided in Table I. The cut-off value of 9% (0.09) was chosen as it equals three times the standard deviation from the mean value of the control (AHA1). Interestingly, the “hit” list reveals novel suppressors involved in Ras/Raf/MEK/Erk and PI3K/Akt signaling pathways, along with the kinases previously identified by our small-molecule kinome inhibitor screen (e.g. FGFR1, Raf) (29), which further validates our approach.

Fig. 1.

Representative hits of the kinome esiRNA screen. Average normalized fluorescence intensity (ΔFI_avg) values of ΔF508-CFTR cells (coexpressing eYFP(H148Q/I152L) that were transfected with esiRNA directed toward A, FGFR1, B. RIPK4, C, MET, D, SHPK, E, MAP3K13, F, BRAF, G, DUSP22, H, CDK10, I, IPMK, or luciferase (nonsilencing control), and grown at 37 °C. After 72 h ΔF508-CFTR cells were stimulated with FIG (25 μm Forskolin, 45 μm IBMX, and 50 μm Genistein) and quenching of YFP fluorescence due to of Cl⁻/I⁻ exchange was quantified by Cellomics KSR Reader (70–300 cells per well). J, Quantitation of rescue (difference in average fluorescence intensity ΔFI_avg) of ΔF508-CFTR at 24 s after adding iodide solution from three independent experiments (a single field per well, 70–300 cells per field). Data are mean ± S.E.

Table I. Results of the esiRNA screen. The top 20 hit genes that displayed a difference in average fluorescence intensity ΔFI_avg (between FIG-stimulated sample and nonsilencing control) of at least 9%.

Gene name	Protein name	Accession No.	Rescue (%)
RIPK4	Receptor-interacting serine/threonine-protein kinase 4	P57078	22
SHPK	Sedoheptulokinase	Q9UHJ6	20
MAP3K13	Mitogen-activated protein kinase kinase kinase 13	O43283	19
FGFR1	Fibroblast growth factor receptor 1	P11362	16
CDK10	Cyclin-dependent kinase 10	Q15131	16
RPS6KC1	Ribosomal protein S6 kinase delta-1	Q96S38	16
PANK4	Pantothenate kinase 4	Q9NVE7	14
DTYMK	Thymidylate kinase	P23919	14
ERN1	Serine/threonine-protein kinase/endoribonuclease IRE1	O75460	14
BRAF	Serine/threonine-protein kinase B-raf	P15056	13
DUSP22	Dual specificity protein phosphatase 22	Q9NRW4	13
IPMK	Inositol polyphosphate multikinase	Q8NFU5	13
PCK2	Phosphoenolpyruvate carboxylase	Q16822	13
CLK3	Dual specificity protein kinase CLK3	P49761	13
MET	Tyrosine-protein kinase Met	P08581	12
CAMK2B	Calcium/calmodulin-dependent protein kinase type II subunit beta	Q13554	12
SOCS1	Suppressor of cytokine signaling 1	O15524	11
NEK10	Serine/threonine-protein kinase Nek10	Q6ZWH5	11
PRKAR2B	cAMP-dependent protein kinase type II-beta regulatory subunit	P31323	9
PANK1	Pantothenate kinase 1	Q8TE04	9

As both the transfection and knockdown efficiency can influence the level of rescue observed in our esiRNA screen, we further validated the hits (and possible off-target effects) by carrying out an independent high-content screen, using shRNA.

Validation of the Hits

To validate the effect of knockdown of the suppressor genes on ΔF508-CFTR chloride channel activity with esiRNA, we tested the top “hits” with shRNA. For this we screened 133 shRNA clones targeting 20 suppressors identified in our kinome esiRNA screen (supplemental Table S2). ΔF508-CFTR cells transfected with shRNA constructs for the suppressor genes or luciferase (nonsilencing control) were analyzed in the Cellomics halide-exchange assay. In parallel, qPCR was performed to determine the knockdown efficiency of these constructs. Although the shRNA clones yielded varied degrees of knockdown, most of them resulted in more than 50% knockdown of target genes (supplemental Table S2), and in general, the degree of ΔF508-CFTR rescue correlated with knockdown efficiency. In the case of MET and BRAF genes, cell death was observed upon knockdown higher than 60–70% and, therefore, shRNA clones that resulted in the best rescue exhibited knockdown of 30% (B-Raf)–60% (MET).

The results of the shRNA screen corroborate the esiRNA hits (Fig. 2 and Table II). Out of 20 analyzed genes, 14, when knocked-down by shRNA, produced reproducible rescue of ΔF508-CFTR function (ΔFI_avg of 12–30%). These included genes encoding the receptor Tyr kinases FGFR1 and MET, as well as B-Raf and the MAPK kinase kinase MAP3K13. Examples of other hits are the receptor interacting protein kinase RIPK4, the sedoheptulokinase SHPK, the cyclin-dependent kinase CDK10, the suppressor of cytokine signaling SOCS1, and the PKA regulatory subunit PRKAR2B. The knockdown of six genes, PANK1, NEK10, CLK3, DTYMK, ERN1, and CAMK2B, led to a lower degree of rescue (ΔFI_avg < 10%) (Table II).

Fig. 2.

Effect of shRNA-mediated knockdown of the suppressor genes on ΔF508-CFTR channel activity. Average normalized fluorescence intensity of ΔF508-CFTR cells transfected with shRNA for A, FGFR1, B, RIPK4, C, MET, D, SHPK, E, MAP3K13, F, BRAF, G, DUSP22, H, CDK10, I, IPMK, or luciferase (nonsilencing control), and grown at 37 °C. After 48 h ΔF508-CFTR cells were subjected to puromycin selection (3 days) and stimulated with FIG (25 μm Forskolin, 45 μm IBMX, and 50 μm Genistein). Quenching of YFP fluorescence during Cl⁻/I⁻ exchange of 70–300 cells per field was recorded and quantified simultaneously by Cellomics ArrayScan VTI. Multiple shRNA clones per gene were analyzed. One representative shRNA clone is shown (Supplemental Table S2). KD, knockdown efficiency (%). J, Quantitation of rescue (difference in average fluorescence intensity ΔFI_avg) of ΔF508-CFTR at 24 s after adding iodide solution. Data are mean ± S.E. from 2–3 independent experiments (three fields per well, 70–300 cells per field). Inset: Comparison of normalized average fluorescent intensity of ΔF508-CFTR (F508-CFTR) versus WT-CFTR.

Table II. Validation of hits with shRNA using the halide-exchange assay Hits were validated by functional assay (Cellomics). Rescue by the best shRNA clone and th.e corresponding knockdown level are shown.

Gene name	Validation by functional assay (Cellomics)		Knockdown level (%)
	Analyzed shRNA clones	Rescue by the best shRNA clone (ΔFIavg)
FGFR1	10	0.30	94
RIPK4	2	0.22	26
MET	17	0.16	61
SHPK	3	0.15	52
MAP3K13	4	0.15	65
BRAF	7	0.15	34
DUSP22	7	0.15	71
CDK10	6	0.14	87
IPMK	10	0.13	63
RPS6KC1	2	0.13	96
PRKAR2B	2	0.12	40
PANK4	15	0.12	80
SOCS1	5	0.12	N/A
PCK2	4	0.12	78
CAMK2B	3	0.09	85
DTYMK	10	0.09	88
ERN1	1	0.08	79
CLK3	8	0.08	64
NEK10	10	0.08	84
PANK1	7	0.06	79

To further validate the hits, we analyzed maturation of ΔF508-CFTR in response to knockdown of the identified suppressors, using immunoblotting for the mature (Band C) protein. supplemental Fig. S1B shows that knockdown of most of the analyzed suppressor genes led to at least 10% increase of band C/B ratio relative to nonsilencing control, except PRKAR2B and DTYMK, where only 7 and 9% increase of band C/B ratio was observed, respectively.

To demonstrate the appearance of ΔF508-CFTR at the plasma membrane, we employed an ELISA assay. For this, we generated a new stable HEK293 GT cell line that expresses ΔF508-CFTR protein with a triple HA tag at the ectodomain (ΔF508-CFTR 3HA cells) (supplemental Fig. S1A). The HA tag was introduced at a site that is known not to affect channel activity, and previously used by another group (22). The results of the ELISA experiments revealed a prominent increase (approx. 50%) in the amount of surface ΔF508-CFTR following knockdown of FGFR1 relative to control knockdown (Fig. 3). Other robust inhibitors (increase in ΔF508-CFTR surface expression by at least 20%) were SHPK, BRAF, CDK10, IPMK, DUSP22, PANK4, SOCS1, and NEK10. The knockdown of six genes (MET, MAP3K13, RPS6KC1, PRKAR2B, PCK2, CAMK2B, DTYMK, CLK3, and PANK1) resulted in a lower degree of rescue (10–20%, Fig. 3).

Fig. 3.

Effect of shRNA-mediated knockdown of the hit genes on surface expression of ΔF508-CFTR. 293MSR-GT cells stably expressing ΔF508-CFTR (bearing a 3HA tag at the ectodomain) were transfected with shRNA for the analyzed genes or nonsilencing control (as indicated), grown at 37 °C for 48 h, selected on puromycin, and quantitation of surface expression of ΔF508-CFTR was carried out by ELISA assay. SU5402 represents cells treated with SU5402 and serves as a positive control. Data are mean ± S.E. from three independent experiments.

In order to measure ΔF508-CFTR channel activity by short-circuit current of our hits, 10 human-to-canine compatible shRNA clones (CDK10, PANK1, PANK4, RPS6KC1, DUSP22, SOCS1, FGFR1, CLK3, NEK10, BRAF, PCK2, and IPMK) were transduced (via lentiviral infection) into MDCK cells stably expressing ΔF508-CFTR (28, 29). Knockdown efficiency was measured by qPCR. PANK1, PANK4, and NEK10 genes showed no expression in the MDCK cells, and the knockdown level of three others (BRAF, PCK2, and SOCS1) was negligible. The remaining genes (CDK10, RPS6KC1, DUSP22, FGFR1, CLK3, and IPMK) exhibited knockdown level of 39–86% (supplemental Table S2), and were subjected to short-circuit current (I_sc) analysis in Ussing chambers. Three of the analyzed genes, RPS6KC1, IPMK, and CLK3, partially restored the ΔF508-CFTR function, as shown by an increase in short-circuit current (21–50%) (supplemental Fig. S2). As MDCK cells exhibited an increased sensitivity toward knockdown of CDK10, DUSP22, and FGFR1 (changes in proliferation rate and/or cell morphology), we were unable to assess ΔF508-CFTR chloride channel activity in the cells that expressed shRNAs for these genes.

FGFR Signaling Plays an Important Role in the Maturation of ΔF508-CFTR

In our previous study (29) we proposed that FGF Receptors (FGFRs), and possibly other Receptor Tyrosine Kinases (RTKs), suppress maturation of the ΔF508 mutant. To further explore the role of FGFRs in the folding and/or trafficking of ΔF508-CFTR, we knocked down FGFR1, 2, and 3, and their select downstream signaling proteins. ΔF508-CFTR cells were transfected with shRNA for FGFRs, Erk1 and 2, Akt, PLCγ-1, FRS2α, or luciferase (nonsilencing control). We then performed the Cellomics halide-exchange assay and immunoblotting analyses. The effect of the knockdown on ΔF508-CFTR maturation and function is shown in Fig. 4. Knockdown of the analyzed genes (i.e. FGFRs, Erk1, Erk2, Akt, PLC-γ1, and FRS2) led to an increase in ΔF508-CFTR channel activity in the halide-exchange assay (ΔFI_avg ≤ 17%), as well as an appearance of band C in the immunoblot (the increase of band C/B ratio of 25–75%), which further confirms our hypothesis that FGFR signaling normally inhibits rescue of ΔF508-CFTR.

Fig. 4.

Correction of the ΔF508-CFTR defect following knockdown of FGF receptors and downstream signaling proteins. Average normalized fluorescence intensity of ΔF508-CFTR cells that were transfected with shRNA for A, FGFR2, B, FGFR3, C, Erk1, D, Erk2, E, Akt, F, PLC-γ1, G, FRS2. H, Quantitation of rescue (difference in average fluorescence intensity ΔFI_avg) of the hits at 24 s after adding iodide solution. Data are mean ± S.E. from three independent experiments, each performed in triplicates (three fields per well, 70–300 cells per field). KD stands for knockdown efficiency (%). I, Immunoblot analysis for maturation of ΔF508-CFTR in 293MSR-GT cells (stably expressing ΔF508-CFTR) transfected with shRNA for FGFR1, 2, and 3, Erk1 and 2, Akt, PLC-γ1, or luciferase (nonsilencing control). 27 °C represents temperature rescue of ΔF508-CFTR at 27 °C, and WT CFTR represents wild-type CFTR. The 27 °C and WT CFTR lanes were loaded with a half of the amount of protein in comparison to the other analyzed samples. (J) Quantitation of rescue (increase in the band C/B ratio) of ΔF508-CFTR, following shRNA-mediated knockdown of the indicated signaling proteins.

Interestingly, inhibition of FGFR1 with the small molecule compound SU5402 administered to ΔF508-CFTR homozygous mice resulted in partial ΔF508-CFTR rescue, as shown by an increase in saliva secretion (supplemental Fig. S3), a surrogate “sweat test” assay in mice (35). As salivary secretion is often sex dependent, only male mice were chosen for these experiments (35). Our results indicate that treatment of the ΔF508-CFTR mice with SU5402 restores the saliva secretion level to ∼10% of that observed for the wild-type CFTR mice, which suggests that indolinone derivatives (such as SU5402) could have therapeutic benefits to CF.

To investigate the effect of FGFR1 inhibition in a highly relevant tissue, we tested the effect of SU5402 on rescue of ΔF508-CFTR function in intestinal organoids (mini-guts) derived from the ileum of ΔF508-CFTR mice (36). As seen in Fig. 5, SU5402 (10 μm) treatment of these organoids resulted in rescue level equivalent to temperature rescue, and ∼85% of WT-CFTR (at 50 min) (Fig. 5C, 5D). This rescue was equivalent or stronger than that observed with VX-809 (3 μm) (Fig. 5D).

Fig. 5.

Rescue of ΔF508-CFTR in intestinal organoids from ΔF508/ΔF508 mice. A–D, Rescue of ΔF508-CFTR with SU5402 analyzed in intestinal organoids from ΔF508/ΔF508 mice by organoids swelling: A and B, Intestinal organoids derived from crypts isolated from the terminal ileum of ΔF508 mice (and WT littermate controls) were treated (at 37 °C) with 5 μm forskolin for 50 min to activate CFTR, leading to swelling via chloride efflux and lumenal fluid accumulation by WT-CFTR but not ΔF508-CFTR. Scale bars = 50 μm. C, ΔF508 organoids do not exhibit increased surface area following forskolin treatment, but swelling can be rescued at low temp: As in B, except one batch of ΔF508 organoids were preincubated at 27 °C for 24 h prior to assay, revealing temp. rescue. D, Treatment with SU5402 or VX-809 partially rescued swelling in ΔF508-CFTR organoids. ΔF508 organoids were pretreated with VX-809 (3 μm; green) or SU-5402 (10 μm; red), administered to the organoid culture media. Treatment with either compound (at 37 °C) partially rescued CFTR-mediated swelling in the ΔF508 organoids, a rescue augmented by combining both SU5402 and VX-809. Error bars for panels B–D represent mean ± S.D. n = 20–30 organoids per treatment. Forsk: Forskolin.

Given the role of FGFR1 suppression in rescue of ΔF508-CFTR, we reasoned that FGFR1 (and its downstream signaling) might inhibit chaperones that promote maturation of ΔF508-CFTR, or promote suppressors of this maturation. To identify these downstream chaperones, we treated ΔF508/ΔF508-CFTR HBE cells from lung explants of two CF patients with 1 or 10 μm SU5402 for 48 h. Our parallel testing of cells from these two patients for SU5402 sensitivity in an Ussing chamber revealed that they exhibit differential sensitivity to this compound at 1 μm; We thus termed these HBEs as responder and poor responder (Fig. 6A). mRNA isolated from these SU5402-treated (or not) cells incubated with chaperone arrays revealed a strikingly different response pattern in the responder versus poor responder (Fig. 6B), underscoring the variability in drug response of patients with the same CF mutation (ΔF508/ΔF508). Analysis of the stronger responder identified several chaperones and regulatory proteins with enhanced expression after treatment (e.g. HSPA 4, 1A, 5, 14, 9; DNAJA2; HSF 1, 4; CRYAB; CCT 3, 5, 7, 6B; HSPE1; SERPINH1), and a few with reduced expression (Bag4; DNAJ C1, C14, C21; TOR1A) (supplemental Table S3). The effect of these chaperones on cell surface expression of ΔF508-CFTR following their overexpression or knockdown with shRNA was then validated using an ELISA assay (Fig. 7A, 7B). Curiously, several of the chaperones exhibiting enhanced expression following SU5402 treatment are known effectors of the transcription factor HSF1 (see below), and accordingly, constitutively active HSF1 led to elevated cell surface expression of ΔF508-CFTR relative to nonactivated HSF1 (supplemental Fig. S4).

Fig. 6.

Chaperone expression analysis following FGFR inhibition by SU5402. A, Response (Isc) of ΔF508/ΔF508-CFTR HBE cells obtained from two separate patients to SU5402 analyzed by Ussing chambers, highlighting the response of the Responder and Poor responder to 48 h treatment with 1 μm or 10 μm SU5402. CFTR currents were stimulated with FIG (25 μm Forskolin, 25 μm IBMX, and 50 μm Genistein) and after the indicated time inhibited using 15 μm GlyH-101 (Gly). Inset: summary of three separate experiments (mean ± S.E.). B, Heat map of chaperone expression profile following 48 h treatment of ΔF508/ΔF508-CFTR HBE cells (from the Responder and Poor responder) with 1 μm or 10 μm SU5402, as indicated. Data are average of three separate arrays (normalized to DMSO control).

Fig. 7.

Validation of chaperone hits. ELISA analysis of cell surface expression of ΔF508-CFTR (relative to luciferase control) expressed in 293-GT cells following, A, overexpression of the top chaperones that exhibited elevated expression following SU5402 treatment in the heat map (Fig. 6), or B, shRNA-mediated knockdown of chaperones that exhibited reduced expression in the heat map following SU5402 treatment. The extent of knockdown (two clones per gene) is shown in Supplemental Table S3. Data in A and B are mean ± S.E., n = three independent experiments.

Additive Effect of FGFR Inhibitors and VX-809 on Rescue of ΔF508-CFTR

Clinical tests of VX-809 administered alone to ΔF508-CFTR patients have not yielded a significant improvement in lung function of these patients (27), although its combination with VX-770 yielded a small improvement (http://investors.vrtx.com/releasedetail.cfm?ReleaseID=856185); however, VX-770 was recently shown to reduce functional expression of VX-809 (37, 38). Thus, we tested the effect of SU5402 (or another FGFR inhibitor, AZD4547) together with VX-809 using the Cellomics assay. Fig. 8 and supplemental Table S4 show that at 1 to 10 μm of both SU5402 (or AZD4547) and VX-809, there was an additive effect on rescue of ΔF508-CFTR, suggesting these compounds act by different mechanisms, an observation also supported by analysis of rescue of ΔF508-CFTR in mouse intestinal organoids (Fig. 5D).

Fig. 8.

Rescue of ΔF508-CFTR in cells treated with VX-809 and FGFR inhibitors. 293-GT cells stably expressing ΔF508-CFTR were treated with the indicated concentrations of VX-809 plus SU5402 or AZD4547 and analyzed for rescue of ΔF508-CFTR using the halide-exchange (Cellomics) assaay. Data are mean ± S.E., n = three independent experiments, each performed in triplicates, three fields per well (70–300 cells per field).

DISCUSSION

Our previous identification, using small-molecule kinase inhibitors, of several signaling cascades (e.g. Ras/Raf/MEK/Erk, Wnt/GSK-3β, PI3K/Akt/mTOR, TAK1/p38) (29) that regulate ΔF508-CFTR trafficking and maturation prompted us to perform a more systematic screen for all human kinases and select associated signaling proteins that may regulate ΔF508-CFTR rescue. We therefore performed an esiRNA human kinome screen using our Cellomics high-content assay to search for suppressors that block maturation of ΔF508-CFTR and hence, when knocked-down, lead to rescue of this mutant. The top hits (20 genes) were validated with another RNAi technology, shRNA, to ensure that the rescue observed in the original esiRNA screen was not caused by off-target effects. In parallel, shRNA constructs were used to test the effect of knockdown of these genes on ΔF508-CFTR maturation and surface expression by immunoblotting and ELISA, respectively. The knockdown of several of the analyzed genes (e.g. FGFR1, SHPK, MAP3K13, DUSP22, CDK10, IPMK, and PANK4) led to a substantial rescue of ΔF508-CFTR activity in the halide-exchange assay, and a corresponding robust increase in cell surface amount of ΔF508-CFTR.

The results of both RNAi screens (esiRNA and shRNA) lend further support to the results obtained in our small-molecule kinase inhibitor screen (29). Several of the hits are kinases that are either direct targets, or are involved in the signaling cascades by the kinase inhibitors discovered in our earlier compound screen. For example, FGFR1 was the target of several ΔF508-CFTR correctors that were previously identified in our kinase inhibitor screen (e.g. SU5402, SU6668, PD173074) (29), and in accord, its knockdown here also led to a substantial rescue of this CFTR mutant.

In this study, we show that knockdown of FGF receptors (FGFR 1, 2, and 3), or their intermediate signaling proteins such as PLC-γ1, Erk1/2, Akt, and FRS2α, can stimulate rescue of ΔF508-CFTR trafficking. Furthermore, our results suggest that FGFR1 inhibitors (e.g. SU5402) may prove useful in treatment of CF: The treatment of the ΔF508-CFTR mice with SU5402 led to the increase of the isoprenaline-stimulated salivary secretion (∼10% of the wild-type CFTR mice level), which corresponds to a partial correction of the ΔF508-CFTR defect, and a strong rescue of ΔF508-CFTR with SU5402 was observed in intestinal organoids harvested from these mice. Given the importance of even a partial (10–25%) rescue of ΔF508-CFTR for improvement in CF patients' health (39, 40), our findings could have significant clinical implications. Moreover, modifications of drug dose, scheduling, mode of administration, and the development of SU5402 more potent analogs, may further improve the efficacy of this (or related) compound for the treatment of CF. In addition, the additive effect of SU5402 and VX-809 on ΔF508-CFTR rescue suggests that these compounds act on different cellular targets and hence drug combination may be useful for future CF treatment.

FGFRs are known to activate the Ras/Raf/MEK/Erk and PI3K/Akt pathways (41–43). In addition to FGFRs, among the RTKs identified by our group was HGF Receptor (Met), another potent regulator of Ras/Raf/MEK/Erk and PI3K/Akt pathways. Nevertheless, how RTKs suppress the rescue of ΔF508 mutant is still unknown, but likely involves inhibition of specific chaperones that normally promote folding and maturation of CFTR or its ΔF508 mutant (e.g. Hsp90 complex), or stimulate inhibitory chaperones/cochaperones and their partner E3 ligases (e.g. Hsp70/CHIP or RMA1 complexes) (28, 44–50). In agreement, Erk1 and Erk2 were shown to inhibit Heat Shock Factor 1 (HSF1) activity and thus to suppress the subsequent expression of heat shock proteins (hsps) (51–55). HSF1 was shown to induce the expression of numerous hsps, for example hsp70 (HSPA1A), hsp60 (HSPD1), hsp40 (DNAJ), hsp27(HSPB2), and CRYAB (HSPB5) (hsp20 family) (56–60), as well as HspE1, HspH1, SERPINH1, and Hsp90AB1 (61); several of these (and other) HSF1 effectors were identified in our current array analysis. In support, hsp70, hspA4 and CRYAB were previously shown by us and others to enhance ΔF508-CFTR maturation (28, 62), as did a constitutively active HSF1. Furthermore, the positive role of HSF1 in the correction of various conformational disease models, including cystic fibrosis, was shown recently (63). Other kinases discovered by the esiRNA screen that are involved in MAPK signaling are B-Raf, MAP3K13, and CaMKIIβ. B-Raf acts downstream of Ras and activates Erk1/2 via MEK kinases (64, 65). MAP3K13 plays a role in the activation of JNK and NFκB (66, 67), and CaMKIIβ (Ca²⁺/calmodulin-dependent protein kinase II beta) was shown to activate TAK1, which in turn leads to phosphorylation of hsp27 (68–70). The phosphorylation of hsp27 promotes switching off the hsp27 folding activity in favor of targeting the substrates for degradation (71–73). In support, hsp27 was identified as a component of the CFTR interactome (48), and has recently been reported to target ΔF508-CFTR for degradation via a SUMO-dependent pathway (74).

The results of our RNAi screen also support a negative role of the NFκB signaling in the maturation of ΔF508-CFTR. We identified RIPK4, a member of the Receptor Interacting Protein Kinase family, which is known to activate NFκB (75). The elevated NFκB-mediated IL8 signaling is one of the main contributors to the chronic hyper-inflammation in the CF lung (76–78). However, how the NFκB pathway inhibits maturation of ΔF508-CFTR needs to be elucidated.

We also identified SOCS1 (Suppressor of Cytokine Signaling 1) as an inhibitor of ΔF508-CFTR rescue. SOCS proteins block STAT phosphorylation by inhibiting Janus kinases (JAKs) activity or competing with STATs for binding sites on cytokine receptors (79–81). In support, we previously demonstrated that overexpression of STAT1 promotes the rescue of ΔF508-CFTR (28).

Another interesting gene identified in our esiRNA screen is PRKAR2B, which encodes the RIIβ regulatory subunit of cAMP-dependent kinase PKA. Stimulation by PKA is known to promote CFTR channel activity, enhance CFTR trafficking, and to decrease CFTR endocytosis from the plasma membrane (82, 83). It was shown that the cAMP-binding domain B of RIIβ subunit inhibits the PKA holoenzyme activation (84). Therefore, the knockdown of this inhibitory subunit could lead to an increase of PKA activity, which in turn promotes ΔF508-CFTR rescue.

Our screen also identified ERN1 (Serine/threonine-protein kinase/endoribonuclease IRE1 or Inositol-requiring protein 1) and IPMK (Inositol Polyphosphate Multikinase) as suppressors of ΔF508-CFTR rescue. ERN1/IRE1 signaling is known for its role in the unfolded protein response in the ER lumen and was reported to reduce the level of misfolded CFTR (85). IPMK is a PI3-kinase that acts as a molecular switch for Akt, inhibiting or stimulating PI3K/Akt signaling (86). PI3K/Akt cascade, along with Ras/Raf/MEK/ERK, is known to affect expression and function of numerous chaperones (87). However, the exact role of IPMK in CFTR maturation and trafficking requires further investigation.

In summary, this study has identified several novel suppressors of ΔF508-CFTR rescue. These suppressors modulate or belong to important cellular signaling pathways such as Ras/Raf/MEK/ERK, PI3K/Akt, p38, and NFκB. The effect of receptor Tyr kinases (especially FGFRs) on ΔF508-CFTR maturation and chaperone expression in ΔF508/ΔF508-CFTR HBE was further elucidated. Thus, identifying proteins and their respective pathways that control rescue of ΔF508-CFTR is a powerful approach to identify drugs that target these same proteins/pathways, and thus can provide new potential treatment avenues for CF.