Research ArticleCystic Fibrosis

Characterizing responses to CFTR-modulating drugs using rectal organoids derived from subjects with cystic fibrosis

See allHide authors and affiliations

Science Translational Medicine  22 Jun 2016:
Vol. 8, Issue 344, pp. 344ra84
DOI: 10.1126/scitranslmed.aad8278

Mini-guts for personalized cystic fibrosis therapy

Cystic fibrosis is caused by mutations in the CFTR gene that severely reduce the function of the CFTR protein. New drugs for treating cystic fibrosis modulate CFTR protein function, but drug efficacy is dependent on which CFTR mutation a patient carries. Dekkers et al. now show that the efficacy of these drugs can be individually assessed in a laboratory test using epithelial cells cultured as mini-guts from rectal biopsies from subjects with cystic fibrosis. The authors show that the drug responses observed in mini-guts or rectal organoids can be used to predict which patients may be potential responders to the drug. This preclinical test may help to quickly identify responders to CFTR-modulating drug therapy even when patients carry very rare CFTR mutations.


Identifying subjects with cystic fibrosis (CF) who may benefit from cystic fibrosis transmembrane conductance regulator (CFTR)–modulating drugs is time-consuming, costly, and especially challenging for individuals with rare uncharacterized CFTR mutations. We studied CFTR function and responses to two drugs—the prototypical CFTR potentiator VX-770 (ivacaftor/KALYDECO) and the CFTR corrector VX-809 (lumacaftor)—in organoid cultures derived from the rectal epithelia of subjects with CF, who expressed a broad range of CFTR mutations. We observed that CFTR residual function and responses to drug therapy depended on both the CFTR mutation and the genetic background of the subjects. In vitro drug responses in rectal organoids positively correlated with published outcome data from clinical trials with VX-809 and VX-770, allowing us to predict from preclinical data the potential for CF patients carrying rare CFTR mutations to respond to drug therapy. We demonstrated proof of principle by selecting two subjects expressing an uncharacterized rare CFTR genotype (G1249R/F508del) who showed clinical responses to treatment with ivacaftor and one subject (F508del/R347P) who showed a limited response to drug therapy both in vitro and in vivo. These data suggest that in vitro measurements of CFTR function in patient-derived rectal organoids may be useful for identifying subjects who would benefit from CFTR-correcting treatment, independent of their CFTR mutation.


Cystic fibrosis (CF) affects about 85,000 persons worldwide and is caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene that encodes an epithelial anion channel (1). Nearly 2000 CFTR mutations have been identified (, which associate with a wide spectrum of phenotypes (, including CF or milder single-organ CFTR-related diseases (2, 3). CFTR mutations are classified into six classes according to their effect on CFTR protein expression and function: (i) no synthesis, (ii) impaired trafficking, (iii) defective channel gating, (iv) altered conductance, (v) reduced amounts of functional CFTR, and (vi) impaired cell surface stability of the protein (1). Although the clinical relevance and CFTR-based treatment options have been well characterized for common CFTR mutations expressed by large groups of subjects, this is not the case for most of the rare orphan mutations expressed by only a few patients ( (2).

Pharmacotherapy targeting the mutant CFTR protein has been successfully developed for a limited number of CFTR mutations. The CFTR potentiator VX-770 (ivacaftor/KALYDECO) has been registered for treating CF caused by G551D (4, 5), S1251N, and seven other CFTR gating mutations (6), which are carried by ~5% of all people with CF. Recent studies indicated that VX-770 combined with the CFTR corrector drug VX-809 (lumacaftor; the combination is termed ORKAMBI) has some beneficial effects on lung function and decreases pulmonary exacerbations in CF subjects homozygous for the most common CFTR mutation F508del (about 45 to 50% of subjects) (7, 8). Conventional clinical trials to identify drug-responsive subjects among about 50% of subjects who express mutations other than CFTR-F508del are costly, time-consuming, and not possible for those with extremely rare CFTR mutations. New cost-effective methods may help to rapidly extend available CFTR-targeting drugs to subjects with CFTR mutations that are currently not registered for treatment.

Here, rectal organoid cultures (9, 10) derived from the rectal epithelia of 71 individuals expressing 28 different CFTR genotypes (see table S1 for an overview of CFTR genotypes and mutation classification) were used to study residual CFTR function and drug-modulated CFTR function. Rectal organoid cultures are three-dimensional (3D) primary stem cell cultures that self-organize into tissue-recapitulating “mini-guts” in vitro that enable the long-term expansion and biobanking of primary patient tissue using defined growth conditions (10). Here, we used a recently established forskolin-induced swelling (FIS) assay (11, 12) and developed a new assay that measures the steady-state lumen area (SLA) of organoids in the absence of forskolin. Data confirmed that the FIS assay was fully dependent on CFTR as previously reported by us and others (1214). The FIS assay enabled sensitive measurements of CFTR function, allowing the study of residual CFTR function and response to CFTR modulators among subjects with CF. The CFTR-dependent SLA assay facilitated the comparison of CFTR function among CF patients and non-CF controls. Rectal organoids were obtained independent of age and with only limited discomfort (15, 16). We correlated in vitro responses by rectal organoids to VX-809 and VX-770 with published clinical trial data and have provided proof of principle that subjects with CF carrying rare CFTR mutations can be selected for treatment with VX-770 on the basis of data from patient-derived rectal organoids.


The SLA of rectal organoids derived from patients with CF and from healthy controls

We observed that rectal organoids from healthy subjects were phenotypically different from rectal organoids derived from subjects expressing two CFTR mutations (which have been established as CF-causing) under standard culture conditions, in the absence of an additional CFTR activator such as forskolin (Fig. 1A). The various mutations are associated with different disease severities (for example, F508del > A455E > R117H-7T) ( We quantitated the SLA by measuring the lumen area as a percentage of total organoid area (Fig. 1B and fig. S1A). The SLA discriminated between healthy controls (fluid-filled/cystic phenotype, 35 to 70%; wild-type/wild-type, 51% ± 10; wild-type/F508del, 47% ± 11; means ± SD) and class I to V mutant CFTR organoids (non- to low fluid-filled/cystic phenotype, 0 to 10%) at the individual level. In CF organoids, the SLA only discriminated between class I to III and class IV and V mutant organoids at the group level [class I to III (1.3 ± 1.4) versus class IV and V (5.0 ± 3.3) (means ± SD); P < 0.0001] (Fig. 1C and fig. S1B).

Fig. 1. The SLA of rectal organoids from CF patients or healthy controls.

(A) Representative confocal images of calcein green–labeled rectal organoids from patients with different CFTR mutations or healthy controls [wild type (WT)] using standard culture conditions. Scale bar, 100 μm. (B) Quantification of the SLA. Recognition of the total lumen area of rectal WT/WT organoids (xy plane) by Volocity imaging software. The luminal surface area (μm2 and %) of the total organoid surface area (100%) is indicated at the bottom of the image. Scale bar, 80 μm. (C) Quantification of the SLA of rectal organoids derived from individuals expressing two CFTR class I to III (n = 35) or class IV to V (n = 18) mutations or from healthy control subjects either expressing WT/WT (n = 12) or F508del/WT CFTR (n = 4). (See table S1 for the classification of different CFTR mutations and fig. S1B for the responses per subject.) Two to five independent experiments were performed for each subject. Data were generated over a period of 2 years. (D) Paired analysis of measurements for the FIS and SLA assays for rectal organoids from a single healthy control and from a CF individual with the F508del/R117H mutation. The data from the control and CF rectal organoids were derived from three independent wells of the same experiment.

Paired analysis of FIS and SLA assays measuring a single healthy control or single F508del/R117H-7T mutant rectal organoid indicated that a large SLA (>40%) was negatively associated with the increase in relative area of rectal organoids upon stimulation with forskolin (R = −0.7206, P < 0.0001) (12). In contrast, FIS assay measurements were unaffected for organoids having SLA values between 0 and 40% (R = 0.0812, P = 0.4569; Fig. 1D). Because healthy control cultures contained many organoids with a large SLA (Fig. 1, C and D, and fig. S1B), FIS rates in healthy control organoids were negatively affected by their SLA phenotype (Fig. 1D and fig. S1C), leading to underestimation of wild-type CFTR function by the FIS assay as indicated by higher FIS for rectal organoid cultures carrying CFTR mutations with high residual function (for example, R117H). Rectal organoids from CF subjects were all within an average SLA phenotype that did not affect FIS (all below ~10%; Fig. 1C and fig. S1B), allowing direct comparison of FIS among rectal organoids from CF subjects. Together, these data indicate that (i) SLA is a CFTR-dependent phenotype that discriminates between healthy controls and subjects with CF, (ii) FIS rates can be compared among CF rectal organoids, and (iii) FIS rates cannot be compared between healthy control and mutant CFTR organoids because of high SLA in healthy control cultures.

Forskolin-induced residual CFTR function in rectal organoids from CF subjects

We assessed the swelling of rectal organoids derived from 37 individuals expressing various class I to V CFTR mutations using eight different forskolin concentrations (0.008 to 5 μM) to maximize the dynamic range of the assay. As described previously (12), rectal organoid swelling was calculated from 60-min time tracings of the surface area increase relative to t = 0 (Fig. 2A). We observed a forskolin dose-dependent increase in swelling that greatly varied among organoids with different CFTR mutations (Fig. 2B). Swelling at 0.8 μM forskolin was used to compare different mutations (Fig. 2C) or individuals with clear detectable residual CFTR function (Fig. 2D), indicating the potential to discriminate between residual CFTR function of organoids expressing different or identical CFTR mutations (Fig. 2, C and D). At 0.8 μM forskolin, the residual CFTR function of organoids derived from individual donors correlated best with paired ex vivo intestinal current measurements (R = 0.5109, P = 0.0055; Fig. 2E) (1719) and the in vivo sweat chloride concentration (R = −0.7011, P < 0.0001; Fig. 2F) (20, 21) compared to the residual CFTR function detected at other forskolin concentrations (Fig. 2B). Swelling of several rectal organoid cultures expressing severe CFTR genotypes (but not 1811+1G>C/1811+1G>C and 166delTA/3120G>A) were most optimally detected at saturating forskolin concentrations (5 μM; Fig. 2B and fig. S2). Comparisons of residual CFTR function as measured by the FIS or SLA assays indicated that the SLA assay less clearly discriminated among individuals (Fig. 2D and fig. S1B) and genotypes (fig. S1D) than did the FIS assay. Some variation in the SLA phenotype appeared independent of CFTR function. For instance, 1811+1G>C/1811+1G>C CF organoids had a relatively large SLA, whereas FIS and thus CFTR function was absent in these cultures (fig. S1, B and C). These data suggest that the SLA assay has a dynamic range at higher CFTR activity compared to the FIS assay (Fig. 2 and fig. S1). Thus, rectal organoid swelling could be used to quantitate residual CFTR function over a large dynamic range that correlated with known in vivo and ex vivo CFTR-dependent markers.

Fig. 2. Residual CFTR function in rectal organoids measured by FIS.

(A) Quantification of the surface area relative to t = 0 (normalized area) of F508del/A455E mutant rectal organoids at different forskolin concentrations averaged from two independent wells. Data are means ± SD. (B) FIS of rectal organoids with various mutations expressed as the absolute area under the curve (AUC) calculated from tracings comparable to (A) (baseline, 100%; t = 60 min). (The data are similar to the residual CFTR function data presented in fig. S4A.) Data are means ± SD. (C and D) Rectal organoid swelling at 0.8 μM forskolin shown as multiple individuals per CFTR genotype (C) or per individual (D). Data are means ± SD. (C) or means ± SEM (D). (E and F) Pearson correlations of intestinal current measurements (ICM) (E) or measurements of sweat chloride concentration (F) versus rectal organoid swelling at 0.8 μM forskolin. ICM values were normalized to healthy control responses to correct for the use of two different protocols. Sweat chloride concentrations were obtained from the Dutch registry database. Each dot represents one individual. (B to D) Data were generated over a period of 1.5 years; ~80% of data were generated during a period of 4 months using a single batch of complete growth medium to limit technical variability. n, number of subjects. Each subject was measured at two to five independent time points in duplicate. The class I mutations included G542X, R1162X (two different subjects), W1282X, DELE2.3, and E60X. lsc, short circuit current.

Pharmacological restoration of CFTR function in rectal organoids

Next, we studied correction of FIS by VX-809 and VX-770 in rectal organoids (see fig. S3 for a detailed analysis). Mutant rectal organoids devoid of any residual function (1811+1G>C/1811+1G>C and 166delTA/3120G>A) did not show drug-induced swelling, indicating expression of mutant CFTR that did not respond to VX-770 and VX-809. All other CFTR genotypes responded in a forskolin dose-dependent manner to the drug treatments, and genotype-specific profiles were observed (examples are shown in Fig. 3A, and all genotypes are in fig. S4A). VX-809 treatment modestly increased FIS for most genotypes (most likely through the CFTR-F508del mutation), but a greater effect in A455E-expressing organoids suggested a potential trafficking defect for this mutant CFTR (Fig. 3A and fig. S4A). Incubation with only VX-770 greatly enhanced FIS in organoids that expressed known class III to V mutations on at least one allele but only modestly enhanced FIS in organoids expressing class I/class II or class II/class II genotypes. Together, VX-809 and VX-770 synergistically increased FIS in organoids homozygous for the F508del mutation or compound heterozygous for F508del and a class I mutation (N1303K, 711-1G>T, R347P, or A455E). In contrast, VX-770 combined with VX-809 only had minor effects on FIS as compared to VX-770 alone for all other genotypes (Fig. 3A and fig. S4A). These data indicated that the CFTR genotype determined the qualitative response to these CFTR-targeting drugs.

Fig. 3. Pharmacological correction of CFTR function in rectal organoids measured by FIS.

(A) Rectal organoids expressing various CFTR mutations were treated as indicated [dimethyl sulfoxide (DMSO), vehicle; VX-809, 3 μM; VX-770, 3 μM] and stimulated with forskolin (see fig. S3 for detailed analysis; these figures are also presented in fig. S4A). Responses at 0.128 μM forskolin (dotted line) were used for correlations with clinical trial data in (B). The different numbers define specific CFTR genotypes and treatments, and they are associated with numbers 1 to 6 of (B) to (D). Data are means ± SD. (B) Data overview of clinical trials with CFTR-correcting treatments in subjects expressing different CFTR mutations. In each clinical trial, results of the most optimal treatment strategy are presented. For the R117H trial, only data from CF subjects aged >18 were used, because subjects aged 6 to 18 had a different mean baseline FEV1 compared to those in the other trials. The numbers correlate with the numbers in (A), (C), and (D). NS, not significant. (C) CFTR modulator–corrected swelling at 0.128 μM forskolin (AUC at t = 60 min) corrected for the DMSO condition presented per CFTR genotype. These DMSO-corrected responses were calculated from data presented in (A) and fig. S4A. The numbers indicate the responses that represent CFTR genotypes and treatments that have been clinically assessed and are associated with the numbers in (A), (B), and (D). The color profile was based on the clinical effectiveness of studies presented in (B). Data are means ± SD. (D) Pearson correlation of drug-corrected rectal organoid swelling [results from (C)] versus lung function increase [results from (B)]. The numbers are associated with (A) to (C). (E) The VX-809 + VX-770–corrected swelling of rectal organoids at 0.128 μM forskolin (AUC at t = 60 min) corrected for the DMSO control presented per individual. These DMSO-corrected responses were calculated from data presented in (A) and fig. S4A. Data are means ± SEM. Data were generated over a period of 1.5 years; ~80% of the data were generated during a period of 4 months using a single batch of complete growth medium to limit technical variability. n, number of subjects. Each subject was measured at two to five independent time points in duplicate. The class I mutations included G542X, R1162X (two different subjects), W1282X, DELE2.3, and E60X.

To identify an optimal approach for estimating in vivo drug efficacy, we assessed correlations between currently available clinical trial data (summarized in Fig. 3B) (6, 22, 23) and drug-induced FIS for rectal organoids quantitated in different ways [per concentration of forskolin or as the AUC of a forskolin dose range; see fig. S4B for DMSO-corrected responses of the clinically assessed CFTR genotypes]. Because drug-treated organoids expressing S1251N or R117H-7T mutations approached maximal FIS rates (AUC, ~3500) at higher forskolin concentrations (Fig. 2A and fig. S4A), their DMSO-corrected drug responses declined at forskolin concentrations higher than 0.128 μM (fig. S4B), leading to underestimation of drug treatment effects. The strongest positive correlation was observed between the response to therapy in organoids at 0.128 μM forskolin (Fig. 3C) and the in vivo absolute change in percent predicted forced expiratory volume in 1 s (FEV1) (R = 0.8706, P = 0.0240; Fig. 3D) compared to the drug-induced FIS detected at other forskolin concentrations. Comparable results were obtained when drug-induced FIS at 0.128 μM forskolin was correlated with the relative change in percent predicted FEV1 (R = 0.8949, P = 0.0402, n = 5) because no relative data were available for the S1251N mutation subgroup (fig. S5, A and B).

On the basis of these data, we constructed a colored reference map to visualize the clinical potential of CFTR-targeting therapy for individuals with CFTR genotypes with unknown in vivo treatment efficacy (Fig. 3C). This map suggested the potential for clinical responses with VX-770 or combined VX-809 + VX-770 treatment for subjects expressing G1249R and TG(13)T(5) mutations, VX-809 + VX-770 for subjects expressing the A455E mutation, and modest potential of VX-809 + VX-770 for the individual who was compound heterozygous for R334W and R746X. Furthermore, CFTR-N1303K, CFTR-711-1G>T, and CFTR-R347P mutations are likely to be nonresponsive, given that the average responses of F508del/N1303K, F508del/711-1G>T, or F508del/R347P rectal organoids were not higher than those of F508del/class I organoids (Fig. 3C). The FIS responses to VX-809 and VX-770 for each individual were variable among donors with different or identical CFTR mutations, indicating that both the CFTR mutations and an individual’s genetic background modified the response to drug therapy (Fig. 3E). Correlations between individual residual CFTR function and response to VX-770 (R = 0.8030, P < 0.0001), VX-809 (R = 0.0113, P = 0.9495), or both (R = 0. 5294, P = 0. 0013) indicated that response to VX-770 correlated most strongly with residual CFTR function (fig. S6).

Compared to the FIS assay, essentially similar data were generated when the forskolin-independent SLA assay was used to measure drug effects after a 24-hour incubation with VX-770, VX-809, or a combination of the two (Fig. 4, A to C). In contrast to the FIS assay, the SLA assay allowed comparisons between CF and healthy control rectal organoids (Fig. 4B). The SLA of drug-treated organoids reached 10 to 120% of that of wild-type healthy control organoids (set at 100%) (Fig. 4B), depending on the CFTR mutations. In contrast, wild-type healthy control organoids reached 140% SLA upon overnight incubation with VX-770. The SLA of F508del/wild-type organoids was ~80% (instead of the expected 50%) of wild-type healthy control organoids, suggesting a nonlinear relation between SLA and CFTR function at higher CFTR activities. This suggested that drug efficacy expressed as a percentage of wild-type control organoids was somewhat overestimated (Fig. 4B). In line with limited induction of FIS by VX-809 alone at lower concentrations of forskolin (Fig. 3A and fig. S4A), VX-809 did not induce SLA in CF organoids of any CFTR genotype in the absence of forskolin; correction of CFTR function was seen with the combined treatment of VX-809 and VX-770 (Fig. 4, B and C). These results indicated that the SLA assay was not suitable for detection of a response to VX-809 monotherapy and suggested that channel activity of VX-809–corrected CFTR required higher 3′,5′-cyclic adenosine monophosphate (cAMP) concentrations as compared to VX-770–potentiated CFTR.

Fig. 4. Pharmacological correction of CFTR function in rectal organoids measured by the SLA assay.

(A) Representative confocal images of calcein green–labeled rectal organoids with or without a 24-hour combination treatment with VX-809 + VX-770 (VX-809, 3 μM; VX-770, 3 μM). Scale bar, 100 μm. (B and C) Quantification of the SLA of rectal organoids incubated for 24 hours with drug treatments as indicated and normalized to the average of the DMSO-treated WT/WT rectal organoids (set at 100%) per experiment (B) or presented as absolute SLA corrected for DMSO (C). The numbers refer to Fig. 3B. Data were generated over a 5-week time frame with a single batch of complete growth medium to limit technical variability. Data are means ± SD. n, number of subjects. Each subject was measured at two to five independent time points in duplicate. The class I mutations include G542X, R1162X, and W1282X. (D) Pearson correlation of the drug-induced SLA versus lung function increase [the numbers refer to Fig. 3B, and the results are in (C)].

The SLA assay recapitulated the relationship between the CFTR genotype and responses to VX-770 or VX-809 + VX-770 as observed by the FIS assay (Fig. 3C versus Fig. 4C). A significant correlation between drug-induced SLA and an increase in absolute lung function was only observed when data for the R117H CFTR mutation (associated with mild CF disease) were excluded (R = 0.9690, P = 0.0065; fig. S5, A and B). When all six CFTR genotypes were included, we observed a trend for correlation (R = 0.7247, P = 0.1032; Fig. 4D and fig. S5, A and B). This may be because the dynamic range of the SLA assay at high CFTR function prevented correction for residual function associated with CFTR-R117H in the absence of treatment. The drug-induced SLA correlated with a relative increase in lung function (R = 0.9172, P = 0.0282, n = 5; fig. S5, A and B).

Thus, the FIS and SLA assays enabled the characterization of drug-corrected CFTR function in organoids derived from donors expressing class I to V mutations. Responses to drug therapy in the FIS and SLA assays and CFTR genotype were associated with published outcome data from clinical trials with CFTR-correcting drugs (68, 22, 23). This suggested that our rectal organoid model could be used to identify drug-responsive individuals with rare CFTR genotypes.

CFTR correction in rectal organoids from patients with two F508del mutations

To further demonstrate that intersubject variability could be consistently measured among donors independent of their CF-causing mutations, we assessed swelling of 10 F508del homozygous rectal organoids in response to forskolin or various β2-agonists using identical assay and culture conditions to minimize the impact of technical variation (Fig. 5 and fig. S7). β2-Agonists activated CFTR via β2-adrenergic receptor stimulation that directly signaled to adenylyl cyclase, the pharmacological target of forskolin (24). We observed subject-specific residual CFTR function and VX-809–induced drug responses upon forskolin or β2-agonist stimulation [Fig. 5, A to C; see fig. S7 (A to C) for responses of the individual experiments and fig. S7D for results with β2-agonists other than salmeterol]. Subject-to-subject variation was more prominent for β2-agonists compared to forskolin, indicating that β2-agonist receptor signaling may be variable between subjects (Fig. 5C and fig. S7D). We observed a strong correlation between forskolin-induced–uncorrected and VX-809–corrected CFTR function (R = 0.8449, P = 0.0029; Fig. 5D). The subject-specific responses were maintained throughout the period of rectal organoid culture for at least 6 months and after 4 months of storage in liquid nitrogen (Fig. 5, E to H, and fig. S7, E to H).

Fig. 5. Long-term stability of swelling in F508del homozygous rectal organoids after forskolin treatment.

(A and B) Representative confocal images (A) or quantification of the surface area increase relative to t = 0 (averaged from three wells) (B) of calcein green–labeled and forskolin-induced rectal organoids derived from two individuals with the F508del homozygous mutation with or without 24 hours of treatment with VX-809 (3 μM). Scale bar, 90 μm. Data are means ± SD. (C) Swelling of rectal organoids induced by forskolin or rectal organoids preincubated for 24 hours with VX-809 and then induced to swell by forskolin or salmeterol. Data are expressed as the absolute AUC calculated from time tracings comparable to (B) (baseline, 100%; t = 60 min). For each experiment, the cultures were assessed simultaneously to limit technical variation. Responses were averaged from three independent experiments performed at weekly intervals [weeks 1 to 3; see fig. S7 (A to C) for the results of the independent experiments]. (D) Pearson correlation for the uncorrected or VX-809 (3 μm)–corrected FIS from (C). (E to H) After the first three experiments [weeks 1 to 3; averages are presented in (C)], two low-forskolin–responding and three high-forskolin–responding rectal organoids were maintained in culture and measured again at weeks 28 and 29, or they were thawed from liquid nitrogen storage at week 20 and measured at weeks 28 and 29. Data on swelling of VX-809–corrected rectal organoids in response to forskolin (E and F) or salmeterol (G and H) are presented per experiment (E to G) or as an average (F to H). The swelling responses are normalized to the average response of the five cultures per experiment (100%). [See fig. S7 (E to H) for the absolute responses.] In (E) and (G), the SD represents the variation from three independent wells; in (F) and (H), the SD represents variation between different experiments. Data are means ± SD.

Together, these data indicated that subject-specific residual CFTR function and response to drug therapy could be detected independent of culture and technical variability in rectal organoids expressing identical CF-causing mutations. This demonstrated the impact of the subject-specific genetic background on the modulation of residual CFTR function and response to drug therapy.

Organoid-based selection of clinical responders to VX-770

Two subjects with the rare CFTR genotype G1249R/F508del were selected for treatment with ivacaftor on the basis of their effective response to VX-770 in rectal organoids derived from these subjects (Figs. 3C and 6A and fig. S4A). The G1249R mutation was described previously as a mutation in exon 20 of the CFTR gene (25), but the functional impact of this mutation has not been described. Upon treatment of the subjects carrying the G1249R mutation with VX-770 (150 mg twice daily) for 4 weeks, the CFTR-dependent marker (i) nasal potential difference (Fig. 6, B to D) and (ii) sweat chloride concentration (Fig. 6, B and E) improved or became normalized. Both subjects showed improved pulmonary function as measured by a decrease in airway resistance (RAW0.5) (Fig. 5F). FEV1 increased by 13% in subject 2, whereas no clear response in FEV1 was observed in subject 1, who had a lower baseline FEV1 (Fig. 6G). Both patients improved with an increase in body weight and better scores on the respiratory domain of the CF-related quality of life questionnaire (Fig. 6B). After a washout period of 4 weeks, airway parameters decreased and CFTR-dependent markers returned to pretreatment levels (Fig. 6, B and D to G). Another subject with the F508del/R347P mutation whose rectal organoids displayed a relatively weak response to VX-770 (Fig. 3C and fig. S4A) showed a limited response to VX-770 in vivo, with a sweat chloride concentration of −14 mM and no improvement in FEV1 (−1%) after 4 weeks of treatment (fig. S8). This suggested that the R347P mutation is not responsive to VX-770 treatment (26). These data indicate that in vitro rectal organoid assays may be able to identify clinical responders to VX-770 even when they express extremely rare uncharacterized CFTR mutations.

Fig. 6. Rectal organoid–based selection of clinical responders to VX-770 treatment.

(A) F508del/G1249R mutant rectal organoids derived from two different donors were treated with DMSO (vehicle), VX-809 (3 μM), VX-770 (3 μM), or VX-809 and VX-770 combined and were stimulated with forskolin. (The average response of both subjects is presented in fig. S4A.) Subjects were measured at five (subject 1) or four (subject 2) independent culture time points in duplicate. Data are means ± SD. (B) Baseline characteristics and VX-770 (ivacaftor/KALYDECO) treatment for two CF subjects expressing the F508del/G1249R mutation. RAW0.5, airway resistance at a flow rate of 0.5 liter/s; NPD, nasal potential difference; CFQ-R, Cystic Fibrosis Questionnaire—Revised; Cl-free, chloride-free; Iso, isoproterenol. (C) Tracings of nasal potential difference measurements before and after treatment of subject 1, performed according to the standard operating procedure of the European Cystic Fibrosis Society Clinical Trial Network (ECFS-CTN). Measurements in the right and left nostril were comparable. Am, amiloride; ATP, adenosine 5′-triphosphate. (D to G) Effects of 4-week VX-770 treatment (KALY) and a 4-week washout period (W) for nasal potential difference (D), sweat chloride concentration (E), RAW0.5 (F), and FEV1 (G).


Here, we have studied subject-specific residual CFTR function and response to CFTR-modulating drugs in rectal organoids from CF patients. The data are most consistent with a model in which residual function and response to therapy are continuous variables (Figs. 2D and 3E), dependent on both the CF-causing mutation and additional subject-specific genetic modifiers. Our data indicate that the current classification model that associates residual CFTR function with only class IV and V mutations may be inaccurate. We suggest that a refinement is needed that more accurately types residual function and pharmacological responses for different CFTR mutations and integrates subject-specific genetic modifiers to indicate the breadth of functional variability for a particular CFTR genotype.

Both the FIS assay and the newly developed SLA assay have specific characteristics. The FIS assay is more sensitive than the SLA assay, and this simple, rapid swelling readout is optimally suited to compare CFTR residual function in CFTR mutant rectal organoids (Fig. 3A and fig. S4A). However, differences in SLA between healthy control and CF rectal organoids prevented direct comparison of these groups (Fig. 1 and fig. S1). The SLA assay has a dynamic window with higher CFTR function and allows for comparisons of drug responses between healthy control and CF organoids (Fig. 4B). However, it is less well suited for measuring residual CFTR function in organoids containing CF-causing mutations (Fig. 1 and fig. S1). Potentially, the SLA readout may also be used to quantitate CFTR function for mutations associated with milder CFTR-related diseases. The SLA assay correlated with measures of in vivo efficacy upon overnight treatment with drug (Fig. 4D and fig. S5) but required a time-consuming quantification method with the current setup and software (Fig. 1B). The SLA and FIS assays together should help to type residual CFTR function and drug-modulated CFTR function for many CFTR mutations.

The relationship between the CFTR genotype and FIS (Fig. 2, B and C) reflects published CFTR genotype-phenotype relationships obtained from clinical registries ( (2729). Mutations 1811+1G>C, 3120G>A, 117-1G>T, N1303K, R347P, and F508del were associated with severe CF and low FIS. Mutations A455E, S1251N, and R334W were associated with milder CF and moderate FIS. Mutations R117H-7T and TG(13)T(5) were associated with CFTR-related diseases and high FIS. Correlations among in vitro organoid responses and established CFTR-dependent markers (sweat chloride concentration and intestinal current measurements) at the individual level (Fig. 2, E and F) further supported our hypothesis that individual residual function measurements in organoids from subjects with CF may have complementary value to current approaches as a diagnostic or prognostic marker for individual patients. The many measurements from organoids compared to sweat chloride concentration measurements in vivo and intestinal current measurements in ex vivo rectal biopsies allow for better control of technical variability and the establishment of forskolin dose ranges that precisely type residual CFTR function. The repeated measurements with organoids and strict CFTR dependency, as supported by analysis of organoids with CFTR null alleles or treated with CFTR inhibitors, in this study and other previous studies (1214), may better indicate subjects who differ in residual CFTR function. To further optimize the detection of small differences in CFTR function among organoids, technical variation may be further minimized by including identical reference organoids on experimental plates. Direct studies that compare FIS levels or other individual markers are needed to compare their value for prediction of individual CF disease severity.

We observed clear differences between residual function and response to therapy among organoids with identical CF-causing mutations that were stable over extended culture periods (28 weeks), different media batches, and independent of biobank sources (Figs. 2D and 3, E and F, and fig. S7). These data extend previous observations demonstrating long-term genomic integrity and epigenetic profile stability of organoid cultures (9, 3032), and point to the individual genetic background as a modifier of CFTR function and fluid transport. Given that organoid swelling is fully CFTR-dependent (1214), variation in swelling observed among organoids derived from different subjects most likely relates to the amount of active apical CFTR channels. This may depend on genetic variability (i) in the CFTR gene itself (because only the CF-causing mutations were characterized) or (ii) in other genomic regions that modify CFTR transcription, translation, posttranslational processing (for example, folding efficacy and protein stability at the plasma membrane), or signaling efficacy between forskolin and CFTR. In addition, non-CFTR factors that modify the driving force for ion transport through CFTR may contribute to subject-to-subject variability, such as (i) function of basolateral and apical ion transporters or channels (14), (ii) aquaporin expression, or (iii) paracellular fluid transport. Subject-specific drug efficacy may further be modulated by genetic variability in drug efflux pumps. These potential mechanisms need to be addressed in more detail in future studies, as well as the relevance of variation in swelling in organoids from subjects with identical CF-causing mutations for clinical outcome.

Our data suggest that CFTR function analysis in rectal organoids may be used to select individuals for treatment with CFTR modulators (Figs. 3 and 5). The established forskolin dose ranges were critical to determine the optimal assay conditions for in vivo correlations. We found the highest positive correlation with clinical responses at a suboptimal forskolin dose (0.128 μM; Fig. 3D). Defining a single forskolin concentration will be particularly helpful when analyzing larger patient populations in follow-up studies to define in vitro–in vivo relationships by limiting the amount of data points. At a forskolin concentration of 0.128 μM, the dynamic range of the FIS assay after drug treatment did not yet reach its maximum for organoids expressing mutations with high residual function such as R117H-7T (Fig. 3A and fig. S4), whereas drug efficacy for CFTR variants with low residual function could be detected. Furthermore, average drug responses of F508del/F508del mutant organoids were about twofold higher than those of F508del/class I mutant organoids (Fig. 3C), suggesting that the assay readout was linear. Whereas prediction of VX-809 and VX-770 drug efficacy in rectal organoids may be optimal at 0.128 μM forskolin, other forskolin doses may be better for establishing correlations with other treatments or clinical parameters (for example, the relationship between sweat chloride concentration and residual CFTR function was optimal at 0.8 μM).

Comparing responses between monotherapy and combination drug treatments requires a note of caution. It is likely that differences between in vivo pharmacokinetics of VX-770 and VX-770 + VX-809 are not fully reflected in vitro. Two drugs may not optimally reach the target tissue in vivo in the way that one drug does. This can result in discrepancies between in vitro and in vivo effects when different treatment modalities are compared. For instance, the responses of A455E organoids to combination treatment and S1251N organoids to VX-770 monotherapy are comparable, but similar effects may not be observed in vivo (Fig. 3C). On the other hand, we do expect that subjects with A455E will show a better response to the combination therapy in vivo than patients with two F508del mutations, as suggested by our in vitro results (Fig. 3C). Thus, analyses of drug responses in rectal organoids and different treatment modalities in vivo may provide insight into differences between pharmacokinetic properties of different treatments.

Proof of principle for preselecting subjects with uncharacterized CFTR mutations for drug therapy using rectal organoid–based assays was demonstrated upon treatment of two subjects with F508del/G1249R mutations with VX-770 (ivacaftor/KALYDECO) (Fig. 6). The subject with a milder baseline pulmonary phenotype showed a clear improvement in FEV1 after treatment. The lack of improvement of FEV1 in the other subject likely resulted from structural damage associated with severe pulmonary disease and fibrosis that may have prevented a FEV1 response to drug therapy (33). The improvement in RAW0.5 was consistent with the curvilinear relation between FEV1 and RAW0.5 (that is, a higher responsiveness of RAW0.5 at low FEV1) (34, 35). Changes in RAW0.5 in the G1249R-carrying subjects appeared relevant because short-term repeatability studies of this marker that indicated variability under identical conditions in patients with stable wheezing disorder (±28%) or healthy controls (±20%) were smaller (36). Positive treatment effects were achieved for other parameters, including two CFTR-dependent markers (sweat chloride concentration and nasal potential difference) (Fig. 6, B to E) that supported the in vivo efficacy of VX-770. Notably, the in vivo treatments for the three subjects in this study were not blinded, which could have introduced an observation bias. In this context, the strong and consistent improvements in objective CFTR markers (sweat chloride concentration and nasal potential difference) were the strongest indications of an in vivo response. For now, these anecdotal data suggest that a relatively simple preclinical test may help in clinical decision-making concerning the application of CFTR modulator therapy in CF patients with rare CFTR mutations.

Thus far, we have been able to recapitulate most observations about responses to CFTR modulators such as VX-809 and VX-770 in primary airway cell cultures (26, 37). Recent data show that chronic VX-770 treatment diminished VX-809–restored CFTR-F508del expression and function in primary airway cells (38, 39). However, we observed with our SLA assay that chronic VX-770 and VX-809 cotreatment was better at restoring CFTR-F508del function than treatment with a single drug (Fig. 4). Differences in experimental conditions between CFTR function measurements in airway cells and rectal organoids may likely contribute to these observations. The SLA assay in rectal organoids is a cumulative steady-state readout without a supraphysiological stimulus that cannot be compared directly with chronic VX-770 stimulation followed by an acute CFTR function measurement in airway cells (38, 39). An increase in CFTR gating and a decrease in CFTR expression by VX-770 treatment could have resulted in a net increase in SLA over the course of the experiment (24 or 48 hours) (Fig. 4). In addition, tissue specificity may also have an impact on the observed differences. For instance, a higher cell turnover rate in rectal organoids compared to differentiated epithelial airway cultures may affect the kinetics of apical CFTR delivery and removal. The down-modulation of VX-809–corrected F508del by chronic VX-770 treatment may be replaced more quickly by newly synthesized CFTR in rectal organoids compared to airway cells and may alter interactions with drugs. Both cell models are relevant and complementary for preclinical drug development and should enable further study of the potential tissue-specific effects of CFTR-targeting compounds.

Several limitations of this study need to be highlighted. Only three subjects were selected for ivacaftor treatment, and larger prospective in vitro–in vivo correlation studies for different CFTR modulators are required to further define the relationship between the preclinical rectal organoid readout and in vivo clinical drug responses. Double-blinded n-of-1 clinical trials with placebo and active drug treatment cycles will help to objectively measure clinical efficacy in individual settings (40) and may help to interpret changes in clinical end points with considerable intrasubject variability such as FEV1 or RAW0.5 (Fig. 6). Longitudinal follow-up is needed to define robust sets of individual markers that are associated with long-term clinical efficacy of CFTR modulators. Combined in vitro and in vivo measurements in double-blind short-term treatment settings will possibly be more informative.

In addition, the 3D organization of rectal organoids that enables the swelling readout also results in limitations. Rectal organoids may rupture upon induction of swelling, although we have only rarely observed this in human organoids over the 60-min time frame of the assay; we have never observed rupture of CF human organoids. Whereas organoid swelling at low swell rates is fully linear over a 60-min time frame, a small curvilinear relation is observed beyond ~30 min for high swell rates (for example, with the R117H mutation), leading to underestimation of CFTR function. Potentially, CFTR function itself may be regulated during cell stretching, although the linear swell response suggests that such an effect does not occur (41). In contrast to electrophysiological measurements with Ussing chambers in 2D systems, selective delivery of compounds to the apical or basolateral compartment in organoids is complicated. Apical stimulation can be performed in organoids by microinjection (42), but this is especially challenging for CF organoids because of their limited luminal volume. Recently described protocols (43) to generate 2D monolayers from 3D organoids in culture may help in the study of ion channel activities that are not coupled to fluid secretion. This would allow for direct comparison of swelling phenotypes and transepithelial current measurements.

In conclusion, here, we have established a relationship between the CFTR genotype, residual CFTR function, and response to therapy using rectal organoids from 71 subjects with CF. We provide a proof of principle that individual in vitro functional measurements in rectal organoids may be used to preclinically select those subjects with CF who will respond to CFTR-modulating drugs. Once rectal organoids are established, they can be biobanked for testing of future drugs and therapeutic combinations in a cost-effective manner. Our data indicate that organoid-based CFTR function measurements can play an important role in the study of rare CFTR mutations and may help to identify subjects with CF who may benefit from CFTR modulator therapy independent of their CFTR mutation.


Study design

Residual CFTR function and responses to the CFTR corrector drug VX-809 (lumacaftor) and CFTR potentiator drug VX-770 (ivacaftor/KALYDECO) in rectal organoids derived from 71 subjects expressing wild-type or mutant CFTR were measured using two assays: the FIS and SLA assays. Quantification of the SLA was performed in a blinded fashion. As proof of principle for organoid-based selection of in vivo drug responders or nonresponders, subjects with a rare CFTR mutation and either a high (F508del/G1249R, n = 2) or low (F508del/R347P, n = 1) organoid response to VX-770 in vitro were assessed in an open-label trial with KALYDECO, dosed according to the manufacturer’s instructions. Four weeks of treatment was followed by a 4-week washout period. Clinical outcome parameters were measured at baseline, after treatment, and after washout. These included the sweat chloride concentration, nasal potential difference, body weight, FEV1, RAW0.5, and Cystic Fibrosis Questionnaire—Revised.

Human material

The ethics committees of the University Medical Center Utrecht and Erasmus Medical Center Rotterdam approved this study, and informed consent was obtained from all participating subjects. Organoids were generated from rectal biopsies after they were used for intestinal current measurements, which were performed for (i) standard care, (ii) voluntary participation in studies, or (iii) diagnosis.

Crypt isolation and organoid culture from rectal suction biopsies

Methods were slightly adapted from protocols described previously (9, 12). In short, crypts were isolated and seeded in 50% Matrigel (growth factor–reduced and phenol-free; BD Biosciences) in 24-well plates (~10 to 30 crypts in three 10-μl Matrigel droplets per well). Growth medium (12) was further supplemented with Primocin (1:500; Invivogen). Vancomycin and gentamicin (both from Sigma) were added during the first week of culture. The medium was refreshed every 2 to 3 days, and organoids were passaged ~1:5 every 7 to 10 days.

The FIS assay

Methods were slightly adapted from protocols described previously (12). In short, rectal organoids (passages 1 to 15) from a 7- to 10-day-old culture were seeded in 96-well culture plates (Nunc) in 5 μl of 50% Matrigel containing 20 to 80 organoids and immersed in 100-μl medium with or without 3 μM VX-809 (Selleck Chemicals LLC). One day after seeding, organoids were incubated for 30 min with 3 μM calcein green (Invitrogen), stimulated with forskolin with or without 3 μM VX-770 (Selleck Chemicals LLC), and directly analyzed by confocal live cell microscopy (LSM710, Zeiss) (everything was performed at 37°C). The total organoid area (xy plane) increase relative to t = 0 of forskolin treatment was quantified using Volocity imaging software (Improvision). Occasionally, cell debris and unviable structures were manually excluded from image analysis based on criteria described in detail in a standard operating procedure. AUC (t = 60 min; baseline, 100%) was calculated using GraphPad Prism.

The SLA assay

Organoids were seeded as described above; incubated for 24 hours with DMSO, 3 μM VX-809, 3 μM VX-770, or their combination; labeled with calcein green; and analyzed by confocal microscopy (all at 37°C). The total area (xy plane) of all organoids in a well was automatically quantified using Volocity, and the luminal area of all organoids of the same well was marked manually and calculated using Volocity in a blinded fashion. The SLA was expressed as the luminal organoid surface area of the total organoid surface area in percentage (Fig. 1B). Occasionally, cell debris and nonviable structures were excluded similarly to the FIS assay.

Intestinal current measurement

Transepithelial chloride secretion in human rectal biopsies was measured using slight adaptations of the procedures described previously (18, 44). See the Supplementary Materials for a detailed description.

Measurement of sweat chloride concentration and nasal potential difference

Both sweat chloride concentration and nasal potential difference measurements were performed according to the most recent version of the standard operating procedure of the ECFS-CTN.

Statistical analysis

Results are presented as means ± SD or means ± SEM, with the number of biological and technical replicates indicated per figure. Statistical analysis was performed by unpaired two-tailed Student’s t test or by Pearson correlation using GraphPad Prism software. The 95% confidence level was considered significant.


Materials and Methods

Fig. S1. Characteristics of the SLA assay in organoids.

Fig. S2. Detectable residual CFTR function in several forskolin-stimulated organoids expressing severe CFTR mutations.

Fig. S3. Quantification of forskolin-induced organoid swelling.

Fig. S4. Genotype-specific profiles of residual and drug-corrected CFTR function in forskolin-stimulated mutant CFTR organoids.

Fig. S5. Data overview of clinical trials with CFTR-repairing treatment in subjects expressing different mutated CFTR genotypes.

Fig. S6. Pearson correlations of residual CFTR function and response to therapy measured by the FIS assay.

Fig. S7. CFTR correction in F508del homozygous organoids.

Fig. S8. In vivo indications for organoid-based selection of a clinical nonresponder to VX-770.

Table S1. Overview of all CFTR genotypes and their corresponding mutation classes and gender.


Acknowledgments: We thank S. Heida-Michel, M. Geerdink, M. C. J. Olling-de Kok (Department of Pediatric Pulmonology, Wilhelmina Children’s Hospital, University Medical Center Utrecht, Utrecht, Netherlands), E. M. Nieuwhof-Stoppelenburg, E. C. van der Wiel (Department of Pediatric Pulmonology, Erasmus University Medical Center/Sophia Children’s Hospital, Rotterdam, Netherlands), J. Hulst, F. T. M. Kokke, B. A. E. de Koning (Department Pediatric Gastroenterology, Erasmus University Medical Center/Sophia Children’s Hospital, Rotterdam, Netherlands), the technicians of the Clinical Chemistry Department (Erasmus University Medical Center/Sophia Children’s Hospital, Rotterdam, Netherlands), N. Adriaens (Department of Respiratory Medicine, Academic Medical Center, Amsterdam, Netherlands), and M. Smink (Department of Pulmonology and Cystic Fibrosis, Haga Teaching Hospital, The Hague, Netherlands) for providing intestinal biopsies, and A. G. M. Bot, M. Rampersad (Department Clinical Chemistry department, Erasmus University Medical Center/Sophia Children’s Hospital, Rotterdam, Netherlands), and M. J. C. Bijvelds (Department of Gastroenterology and Hepatology, Erasmus University Medical Center/Sophia Children’s Hospital, Rotterdam, the Netherlands) for performing intestinal current measurements. Funding: This work was supported by grants of the Dutch Cystic Fibrosis Foundation as part of the HIT-CF program, the Wilhelmina Children’s Hospital Foundation, and the Dutch Health Organization ZonMw, Netherlands. Author contributions: J.F.D. designed and performed the experiments and generated article text and figures. G.B. performed the experiments and generated article text and figures. E.K. and A.V. performed the experiments. H.R.d.J. interpreted the experiments and revised the manuscript. H.M.J., E.A.v.d.G., R.H.J.H., F.P.V., J.C.E., C.J.M., and H.G.M.H. provided patient material. I.B. provided patient material and ICM data. E.E.S.N. provided funding, interpretation of the experiments, and revised the manuscript. Y.B.d.R. provided ICM data. K.M.d.W.d.G. provided patient material and clinical data. H.C. interpreted data and revised the manuscript. C.K.v.d.E. provided patient material, clinical data, interpreted data, and revised the manuscript. J.M.B. designed and interpreted the experiments, wrote article text, and provided funding. Competing interests: J.M.B. received travel funds and speaker honoraria from Vertex Pharmaceuticals, Novartis, and Pfizer. J.M.B., C.K.E., J.F.D., and H.C. are inventors on a patent application related to these findings: “A rapid quantitative assay to measure CFTR function in a primary intestinal culture model” (#WO 2013093812 A2). H.C. is an inventor on other patents related to these findings: “Culture medium for epithelial stem cells and organoids comprising said stem cells” (#WO 2010090513 A2) and “Improved culture method for organoids” (#WO 2015173425 A1).

Stay Connected to Science Translational Medicine

Navigate This Article