Global assessment of the integrated stress response in CF patient-derived airway and intestinal tissues

Abstract

Background:

Chronic inflammation is a hallmark among patients with cystic fibrosis (CF). We explored whether mutation-induced (F508del) misfolding of the cystic fibrosis transmembrane conductance regulator (CFTR), and/or secondary colonization with opportunistic pathogens, activate tissue remodeling and innate immune response drivers.

Methods:

Using RNA-seq to interrogate global gene expression profiles, we analyzed stress response signaling cascades in primary human bronchial epithelia (HBE) and intestinal organoids.

Results:

Primary HBE acquired from CF patients with advanced disease and prolonged exposure to pathogenic microorganisms display a clear molecular signature of activated tissue remodeling pathways, unfolded protein response (UPR), and chronic inflammation. Furthermore, CFTR misfolding induces inflammatory signaling cascades in F508del patient-derived organoids from both the distal small intestine and colon.

Conclusion:

Despite the small patient cohort size, this proof-of-principle study supports the use of RNA-seq as a means to both identify CF-specific signaling profiles in various tissues and evaluate disease heterogeneity. Our global transcriptomic data is a useful resource for the CF research community for analyzing other gene expression sets influencing CF disease signature but also transcriptionally contributing to CF heterogeneity.

1. Introduction

Chronic inflammation is a hallmark clinical manifestation among patients with cystic fibrosis (CF), a lethal autosomal recessive disorder caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene. The most prevalent CFTR defect is a deletion of the phenylalanine at position 508 (F508del), which abrogates CFTR folding and results in a loss-of-function phenotype [1–3]. Lack of CFTR – a chloride and bicarbonate channel – alters electrolyte homeostasis and leads to the accumulation of thick, dehydrated and viscous mucus on the surface of secretory epithelia. This creates a favorable environment for colonization and chronic persistence of opportunistic pathogens such as Pseudomonas aeruginosa, Burkholderia cepacia and Aspergillus fumigatus, which are primary causes of lung morbidity and mortality in most individuals with the disease [4,5]. Long-term bacterial infections and massive inflammatory responses commonly contribute to CF pathogenesis. Shortly after birth, persistent neutrophilic inflammation of small airways is observed, which progresses towards lung injury and pronounced functional defects [6]. Previous reports suggest that structural airway wall changes and tissue remodeling are either caused by prolonged pulmonary inflammation and secondary infection(s) [7], or simply occur in parallel [8]. In the context of CF, it remains unknown whether CFTR misfolding and/or secondary infection(s) trigger the inflammatory response, and as a secondary reaction tissue remodeling. Alternatively, both events could be concordant and interlinked.

In addition to chloride and bicarbonate, CFTR also transports thiocyanate ions, which are key regulators of innate immunity [4]. The load of destabilized and misfolded CFTR protein may impose imbalance within the endoplasmic reticulum (ER), otherwise known as ‘ER stress’, which activates an unfolded protein response (UPR). The UPR reflects an elaborate series of signaling pathways that orchestrate restoration of cellular proteostasis [9] and is deeply connected with inflammation [10]. Some studies suggest that mutation-induced CFTR misfolding (e.g. F508del) elicits ER stress in overexpression systems such as Calu3 and HeLa cell lines [11,12]. Other reports have argued that aberrant CFTR synthesis directly activates an array of immune response drivers [4,13]. Infants with CF begin to exhibit evidence of inflammation within the first few months of life, while being simultaneously exposed to external microflora and opportunistic pathogens. CF ferret models treated with broad-spectrum antibiotics lack detectable bacteria or fungi yet develop neutrophil-dominated inflammation, suggesting an independent contribution of mucoinflammatory processes to disease pathogenesis, even in the absence of overt infection [14]. Thus, it is not trivial to distinguish whether the inflammation is activated initially by the bacterial infection or by the effect(s) of CFTR misfolding.

In the present study, we utilize global transcriptome analyses to elucidate a molecular profile of inflammation and stress response in primary airway epithelia from three CF patients with severe pulmonary exacerbation and one non-CF individual (i.e. disease control), as well as CF patient-derived intestinal organoids collected from an infection-free environment. Through this approach, we address an important and complex question using (for the first time) primary human tissues and leading-edge organoids together with genome-wide transcriptomics. In proof-of-concept fashion, our analyses highlight usefulness of a global, transcriptome-wide approach for comparing stress response signaling patterns in patient-derived respiratory and gastrointestinal materials, in order to: (1) capture multi-layered effects associated with both F508del-induced CFTR misfolding and prolonged exposure to opportunistic human pathogens, and (2), interrogate whether the combination of chronic CFTR misprocessing and/or microbial infection(s) activate signaling cascades involved in tissue remodeling.

2. Methods

2.1. Samples

After obtaining informed consent, primary HBE were isolated from lung tissues harvested by pulmonary transplantation through the Human Airway Cell Core at the University of Pittsburgh (Pennsylvania, USA) under a protocol approved by the University of Pittsburgh Investigational Review Board. HBE were derived from the following three CF patients: female, 45.3 years, CFTR^{F508del/F508del} genotype (code #192); female, 36.5 years, CFTR^{F508del/F508del} genotype (code #209); female, 37.2 years, CFTR^{F508del/G542X} genotype (code #222). In addition, HBE were also collected from a non-CF individual with polymyositis and associated pulmonary hypertension/fibrosis (male, 45 years, code #923). Primary HBE were cultured on transwell filters at 37 °C and 5% CO₂ [15,16], trypsinized, total RNA was extracted using TRIzol reagent (Sigma Aldrich) and then sequenced.

2.2. Organoid culture

Ileal and colonic (rectal) crypts were freshly isolated from intestinal biopsies – from one F508del homozygous CF patient and one healthy individual – by ethylenediaminetetraacetic acid (EDTA) treatment and seeded in Matrigel (growth factor reduced, phenol free; BD Bioscience) as described in detail previously [17]. The culture medium for subsequent organoid expansion contained advanced DMEM/F12 (Invitrogen), supplemented with Glutamax (2 mmol/L; ThermoFisher), murine epidermal growth factor (50 μg/L), cell culture supplements B27 and N2 (all from Invitrogen), [Leu-15]-Gastrin (10 nmol/L), N-acetylcysteine (1 mmol/L), nicotinamide (10 mmol/L), HEPES (10 mmol/L), CHIR99021 (10 μmol/L; only present during the first 2 days after (re-)seeding), SB202190 (3 μmol/L; all from Sigma-Aldrich), Y-27,632 (10 μmol/L; only present during the first 2 days after (re-)seeding), prostaglandin E2 (10 nmol/L), A83-01 (0.5 μmol/L; all from Tocris), and an antibiotic formulation (Primocin; Invivogen). In addition, conditioned cell culture media containing the following growth factors were added: noggin (10% v/v), R-spondin 1 (20% v/v) and Wnt3a (50% v/v). Harvesting of conditioned media from supportive cell lines was performed as described [18]. Media were refreshed every 2–3 days, and organoids were collected and re-seeded at a 1:4 dilution on a weekly basis.

Goblet cell-enriched colonoids were produced by retrieval of Wnt3a, including addition of the Wnt3a secretion inhibitor IWP-2 (10 μmol/L) and the γ-secretase inhibitor DAPT (10 μmol/L; both from Tocris) 7 days after seeding [19]. Goblet cell-enriched organoids were collected after culturing for an additional 4 days. Passage numbers for colonic and ileal organoids were 8 and 7, respectively.

2.3. Cell culture

CFBE41o⁻ cells stably transduced with wild-type or F508del-CFTR [20,21] were grown in Earles MEM (Gibco) supplemented with 10% FBS (Gibco) and 1% glutamine (Gibco), and cultured at 37 °C in humidified atmosphere with 5% CO₂ on dishes coated with collagen (Corning) or human fibronectin (ThermoFisher).

2.4. RNA-seq

Total RNA from each sample was extracted in TRIzol reagent (Sigma Aldrich), DNaseI treated (5 U DNaseI / 5 μg RNA, 30 min, 37 °C), spiked-in, rRNA depleted (Ribo-Zero™ Magnetic Gold kit Human/Mouse/Rat; Illumina), and concentrated (RNA Clean and Concentrator™–5; Zymo Research). The cDNA library generation was performed as previously [22] and sequenced on a HiSeq2500 platform (Illumina).

2.5. Data analysis and statistics

Processed reads were uniquely mapped to the human genome (GRCh37) using Bowtie (0.12.9), allowing a maximum of 2 mismatches (parameter settings: −l 16 −n 1 −e 50 −m 1 — strata — best y). Counts were normalized as reads per kilobase per million mapped reads (RPKM) [23]. Genes listed as downstream targets of activating transcription factor (ATF) 4, ATF6, X-box binding protein 1 (XBP1) and nuclear factor kappa-light-chain-enhancer of activated B cells (NFkB) were extracted from previously published data sets [24–27]. Lists of genes activated by hypoxia-inducible factor 1-alpha (HIF1A), cAMP response element binding protein (CREB) and serum response factor (SRF) were taken from [28]. Fold-change analysis of referenced genes was compared as specified in each analysis. Transcripts exhibiting a fold-change ≥ 2 (log₂ ≥ 1) were designated as upregulated and evaluated in statistical analyses. For upregulated transcripts, increased mRNA levels were considered correlative with protein production at steady-state (i.e. the timepoint at which samples were collected) [29], which reflects transcriptional regulation. In contrast, downregulated transcripts were not associated with decreased gene product levels due to substantial differences in half-lives for individual mRNAs and their encoded proteins [30,31]. P-values less than 0.05 were considered statistically significant as defined by Wilcoxon rank sum test.

2.6. Data deposition

Sequencing data for primary HBE and intestinal organoids have been deposited within Gene Expression Omnibus (GEO) under accession number GSE143621. Sequences from CFBE stably expressing wild-type- or F508del-CFTR are accessible in GEO under accession number GSE74365 [22]. All expression and fold-change values, along with gene IDs and names (e.g. summarized for Figs. 1 & 2 and Suppl. Fig. S1–S4), are listed in the Supplementary Table.

Fig. 1.

Tissue remodeling in primary HBE cells obtained from three CF patients and one non-CF individual.

(A) Schematic of EGF-induced activation of the MAP-kinase and mTOR signaling pathways. Key transcription factors are highlighted (yellow), and the total number (n) of downstream genes they activate are designated. (B-D) Fold changes of genes induced by SRF (B), HIF1A (C) and CREB (D) in patient codes #209, #222, #192 (all CF) and #923 (non-CF). Values obtained from these donors were compared to CFBE stably transduced with wild-type CFTR. For each transcription factor, the number of downstream genes detected (out of the total for each signaling pathway, panel A) is given in the upper left corner. Different sizes of gene sets are noted for HIF1A, CREB and SRF within each patient (B-D) compared to the maximum number of total genes (A), which captures variation in sequencing detection levels for distinct patient-derived materials. The number of transcripts upregulated more than 2-fold is italicized in the upper right corner. The vertical dashed line denotes the 2-fold change threshold. P-values for upregulated targets are defined with Wilcoxon rank sum test.

Fig. 2.

Tissue remodeling in F508del-CFTR intestinal organoids.

(A-C) Fold changes of downstream genes induced by the SRF (A), HIF1A (B) and CREB (C) transcription factors in organoids harvested from the ileum, undifferentiated colon and differentiated (goblet-cell enriched) colon of a CF patient with CFTR^{F508del/F508del} genotype. Values obtained for each type of CF organoid are compared to corresponding organoids generated from a healthy donor. For each transcription factor, the total number (n) of downstream genes detected is given in the upper left corner; the total number of genes in each signaling pathway is shown in Fig. 1A. Different sizes of total gene sets are noted for HIF1A, CREB and SRF within each type of organoid (A-C) compared to the maximum number of total genes (Fig. 1A), which captures variation in sequencing detection levels for distinct intestinal sections. The number of transcripts upregulated more than 2-fold is italicized in the upper right corner. The vertical dashed line denotes the 2-fold change threshold. P-values for upregulated transcripts are defined with Wilcoxon rank sum test.

3. Results

3.1. Tissue remodeling is present in primary CF airway epithelia

Epidermal growth factor (EGF) binds the EGF receptor (EGFR) at the plasma membrane and initially triggers the MAP-kinase pathway, followed thereafter by an elevation of mammalian target of rapamycin (mTOR) [32]. These two signaling cascades primarily augment three transcription factors – hypoxia induced factor 1A (HIF1A), cAMP responsive element binding (CREB) protein, and serum response factor (SRF) – each of which leads to transcription of specific gene sets involved in tissue remodeling [28] (Fig. 1A). To assess potential effect(s) of CF pathology on signaling cascades that modulate tissue remodeling, we performed RNA-seq on primary, non-passaged HBE cells extracted from three CF donor lungs (#192, F508del/F508del; #209, F508del/F508del; #222, F508del/G542X). We utilized HBE freshly obtained from an age-matched, non-CF donor (#923, polymyositis) as a means to define disease-specific signatures. Although the non-CF patient was without any known infections, collateral tissue remodeling has been described in individuals with polymyositis [33], which challenges the suitability of this patient as an appropriate negative control. Therefore, we employed CF bronchial epithelial cells (CFBE) stably transduced with wild-type CFTR to establish a baseline for gene expression comparisons. Comparison between CFBE expressing F508del or wild-type CFTR revealed no evidence of tissue remodeling (Suppl. Fig. 1), implying that these cells are robust and suitable reference for evaluating signaling cascades in this context.

Analysis of the primary HBE from all CF donors revealed consistent transcriptional upregulation of key tissue remodeling pathways, e.g. MAP-kinase (SRF/CREB-activated) targets (Fig. 1). In particular, downstream genes stimulated by SRF and CREB were significantly increased among the three CF patients, with the exception of SRF targets in code #192 (Fig. 1B, D; single gene lists in Suppl. Table 1). Transcripts controlled by HIF1A exhibited a more heterogeneous pattern of activation among all HBE samples, with some sets appearing upregulated but scoring insignificant (Fig. 1C; single gene lists in Suppl. Table 1). Notably, expression of some genes controlled by CREB, SRF or HIF1A was 64-fold higher as compared to baseline levels in CFBE cells (Suppl. Table 1). Although we detected seeming upregulation of SRF, HIF1A or CREB targets in the non-CF patient (#923), statistical tests resulted in insignificance (Fig. 1B–D). Together, our results show strong evidence for CF-triggered activation of MAP-kinase signaling cascades involving genes downstream of the HIF1A, CREB and SRF transcription factors, which are clear indicators of tissue remodeling. Importantly, these findings were only observed in CF airway epithelia and not in polymyositis HBE.

3.2. EGF-induced tissue remodeling is not activated in CF patient-derived intestinal organoids

In primary CF HBE cells, unique molecular signatures associated with tissue remodeling may potentially originate from variations in exposure to microbial pathogens, levels of misfolded F508del-CFTR protein, and/or other factors. Thus, we next investigated whether endogenous F508del expression alone is able to initiate tissue remodeling in CF intestinal organoids derived from an infection-free environment. We generated organoids from a CF patient homozygous for F508del-CFTR and a healthy donor, then compared transcript expression patterns in samples produced from both the distal small intestine (ileum) and colon at two stages of growth: undifferentiated and goblet cell-enriched. In F508del-expressing organoids, we detected almost no changes in both MAP-kinase and mTOR pathways (Fig. 2), with exceptions including modest increases in SRF and HIF1A targets within the ileum (Fig. 2A, D) and CREB-regulated genes in the undifferentiated colon (Fig. 2C). These results indicate near absence of tissue remodeling in CF organoids.

3.3. Heterogeneous activation of the integrated stress response occurs in CF lung epithelia and gastrointestinal organoids

Because individuals with F508del experience life-long CFTR misfolding and exposure to microbial pathogens, we next assessed effect(s) of CF pathology on inter-related branches of the integrated stress response and chronic inflammation. In both primary HBE and CF patient-derived organoids, we compared expression levels of marker genes stimulated by the three main ER stress signaling cascades: (1) protein kinase RNA (PRK)-like ER kinase (PERK), ATF6 and (3) inositol-requiring enzyme-1 (IRE1) (see supplementary text and Suppl. Fig. 2A). In all CF airway epithelia, we observed significant upregulation of genes activated by ATF4, ATF6 and NFkB (Suppl. Fig. 3A–D). In the polymyositis patient, however, only ATF4-gene targets were increased (Suppl. Fig. 3A), suggestive of an acute stress response.

Ileal organoids derived from an F508del homozygous patient displayed the strongest induction of UPR and inflammatory response markers (see supplementary text and Suppl. Fig. 4A–D). All organoid types (e.g. ileal, undifferentiated colon, goblet-enriched colon) exhibited robust upregulation of NFkB-activated genes (Suppl. Fig. 4D), which supports the hypothesis that acute or chronic F508del-CFTR misfolding stimulates inflammatory responses. These findings demonstrate that CF lung tissue from individuals with advanced pulmonary disease, as well as intestinal organoids from an infection-free CF patient, display molecular signatures consistent with activated UPR and chronic inflammation.

Taken together, our results show evidence for F508del-triggered chronic inflammation in three types of intestinal organoids – as judged by NFkB-activated gene expression (Suppl. Fig. 4D) – in a manner that would not be expected to confer tissue remodeling. In contrast, clear indicators of tissue remodeling were observed in CF lung epithelia, as evidenced by strong activation of MAP-kinase and mTOR signaling cascades involving downstream targets of the HIF1A, CREB and SRF transcription factors.

4. Discussion

We address a longstanding question concerning whether activation of innate immunity in CF airway epithelia results from F508del-induced CFTR misfolding or reflects an effect of secondary colonization by opportunistic human pathogens, and in turn, stimulates tissue remodeling cascades. We systematically analyzed the global integrated stress response and EGF-triggered remodeling signals in primary HBE cells collected from multiple CF donors (who harbor recurring pulmonary infections) and intestinal organoids harvested from an individual with CFTR^{F508del/F508del} genotype. This allowed us to interrogate effect(s) of chronic CFTR misfolding on tissue remodeling patterns in patient-derived respiratory and gastrointestinal samples. Aberrant CFTR biogenesis was shown to activate UPR in both lung epithelia and intestinal organoids. In the context of the airway, microbial pathogens appear to also stimulate signaling cascades involved in tissue remodeling. In gastrointestinal organoids, clear signatures of F508del-induced UPR and chronic inflammation were observed, while a significant tissue remodeling response was absent. These findings support the hypothesis that during CF disease progression, tissue remodeling develops as a concordant process irrespective of CFTR misfolding.

Cellular stress and immune response are often activated in an oscillatory manner to adequately balance stress responses and alleviate harmful, chronic signaling cascades [11,12]. These fluctuations might explain the lack of ER stress detection in certain studies. Moreover, stress response pathways initiate at varying time-scales and often in an inversely correlated fashion – i.e. one pathway may decline while others enhance, with different components of the innate response dwelling between active and inactive [34]. A prevalent molecular signature in native CF tissues and organoids is augmented PERK activation (ATF4), causing a strong stress response through transcriptional upregulation of pro-inflammatory cytokines. UPR activation in the organoid model is likely induced by F508del misfolding, or potentially by a combination of F508del-CFTR and long-term epigenetic changes inherent to the CF condition. In primary CF airway epithelia, the inflammation seems to be a cumulative consequence of F508del-induced CFTR misprocessing and colonization with bacterial pathogens, which could explain regularly observed hyper-inflammation in CF tissues. This phenotype is characterized by enhanced neutrophil recruitment and strong induction of pro-inflammatory cytokines [35,36], ultimately culminating in tissue damage and remodeling [8]. CF patient lungs experience recurrent cycles of bacterial infection and inflammation, which elicits degeneration and loss of respiratory function. Lung morbidity is due to reduced numbers of differentiated cells and increased levels of collagen, rendering the tissue less flexible and prone to bronchiectasis [37].

In summary, we utilize global RNA-seq analyses to elucidate a complex pathological question in CF patient-derived respiratory and gastrointestinal tissues. Our study demonstrates in initial, proof-of-principle fashion the ability of transcriptomics to detect differences in stress response signatures within primary tissues and evaluate phenotypic heterogeneity among individuals living with CF. In the future, this technology could be implemented to uniquely characterize CF disease progression among specific age groups, genders, racial/ethnic backgrounds, genotypes, etc., and potentially inform personalized approaches to clinical intervention. Hence, our global transcriptomic data from different CF-patient originating tissues could be a useful resource for the CF research community to select other gene expression sets influencing CF disease signature but also contribute to CF heterogeneity.