Introduction
Viral and bacterial respiratory infections are associated with the majority of exacerbations in chronic inflammatory lung diseases, such as chronic obstructive pulmonary disease (COPD) and asthma. Human rhinovirus (RV) is one of the most common viruses detected during exacerbations in patients with COPD and asthma [
1]. Despite these associations, the pathogenesis of virus-induced exacerbations is incompletely understood, limiting the options for developing therapeutic strategies aiming at reduction of virus-induced exacerbations and thereby disease progression. The airway epithelium that lines human airways is the main target for initial RV infection. Airway epithelial cells are essential for host defense of the lungs. They provide a physical barrier, mount innate immune responses, mediate mucociliary clearance and orchestrate adaptive immune responses [
2].
Airway epithelial cell cultures have been widely used to study RV infections as well as cigarette smoke (CS) exposures, the latter being the primary risk factor for development and progression of COPD. Studies have employed combinations of CS extract with RV exposure in submerged airway epithelial cells and showed impairment of anti-viral defenses with a subsequent increase in viral infection [
3]. Another study used well-differentiated primary bronchial epithelial cells (PBEC) to study the effect of whole CS on RV infections and showed decreased gene expression of interferons and interferon response genes, and inhibitory effects of α-1 antitrypsin treatment [
4]. However, it is currently unknown how the kinetics of early and late antiviral defenses are impaired by CS. In addition, other cellular host mechanisms induced by CS exposure that contribute to the increased viral load of CS-exposed airway epithelium are currently unexplored. In addition, studies have indicated that COPD patients may be more susceptible to RV infection [
5], but it is not currently known how this is mediated.
To address the above, we used air–liquid interface (ALI)-cultures of well-differentiated PBEC from COPD and non-COPD donors that adequately represent the human airway epithelium in situ [
6], combined with exposure to whole CS (gas and particulate phase) [
7]. Using an unbiased RNA sequencing (RNA-Seq) approach, we aimed to explore signaling pathways regulated by CS exposure or RV infection, or their combination, and delineated possible mechanisms involved in the effect of CS exposure on RV replication. Furthermore, we performed cellular deconvolution analysis to identify possible changes in cellular composition in ALI-PBEC upon the different exposures. Importantly, we validated these findings in RV-A16-infected ALI-PBEC by comparing these to those obtained by analysis of a human RV-A16 challenge model. These findings provide further insight into the pathogenesis of virus-induced exacerbations and may help to explain higher RV infection levels of CS-exposed airway epithelium.
Materials and methods
Subjects
PBEC were isolated from tumor free, macroscopically normal resected lung tissue (bronchial rings) from patients undergoing surgery for lung cancer at the Leiden University Medical Center from COPD (GOLD stage II) donors and non-COPD controls, as described previously [
8]. For CS and RV16 exposure experiments, 8 COPD and 8 non-COPD donors were included and viral load measurements and qPCR analysis were performed. From these donors, 7 COPD and 6 non-COPD donors were included for RNA-sequencing analysis; cells from one COPD and two non-COPD donors were excluded in the RNA-Seq analysis based on lower viral RNA levels in air-exposed ALI-PBEC. The experiments on interferon (n = 4), FCCP (n = 5),
l-lactate (n = 3) treatments and GDF15 knockdown (n = 4) included only non-COPD donors. The baseline characteristics of COPD and non-COPD donors included for all experiments in this study, are provided in Table
1.
Table 1
Baseline characteristics of COPD and non-COPD donors
Subjects (n) | 8 | 8 | 7 | 6 | 4 | 5 | 3 | 4 |
Sex ratio (Female/Male) | 3/5 | 4/4 | 2/5 | 2/4 | 3/1 | 1/4 | 1/2 | 2/2 |
Smoke status | Ex-smoker (8) | Ex-smoker (8) | Ex-smoker (7) | Ex-smoker (6) | Ex-smoker (1) non-smoker (3) | Ex-smoker (2) non-smoker (2) Unknown (1) | Smoker (1) ex-smoker (1) non-smoker (1) | Smoker (1) ex-smoker (3) |
Age (years)# | 63.0 ± 2.6 | 65.4 ± 3.0 | 63.7 ± 2.8 | 64.3 ± 3.6 | 53.8 ± 3.1 | 58.2 ± 1.9 | 64.0 ± 1.2 | 68.3 ± 2.8 |
FEV1 (%predicted)# | 65.6 ± 4.0 | 101.9 ± 6.7* | 64.4 ± 4.4 | 99.7 ± 6.2* | 103.0 ± 3.6 | 96.0 ± 3.5 | 101.3 ± 16.7 | 113.1 ± 6.9 |
FEV1 (L)# | 1.9 ± 0.1 | 2.7 ± 0.2* | 1.9 ± 0.1 | 2.9 ± 0.3* | 3.3 ± 0.2 | 3.6 ± 0.3 | 3.0 ± 0.5 | 2.9 ± 0.2 |
Exposure of primary bronchial epithelial cells
PBEC at passage 2 were seeded at a density of 40,000 cells/insert on 0.4 μm pore size 12-well Transwells (Corning Costar, Cambridge, USA). When confluent, cells were apically exposed to air for 4 weeks to initiate cell differentiation. ALI-PBEC from non-COPD donors (n = 8) and COPD donors (n = 8) were either exposed to whole CS from one cigarette or to room air as control for 4- 5 min, as previously reported [
9]. Following exposure, smoke was removed by ventilation with air for 10 min. Based on the weight of the outlet filter before and after CS exposure, the box in which the 12-well plate was exposed received ~ 2.8 mg of CS particles. Following CS exposure, ALI-cultures were immediately infected by apical exposure with 200 µl of RV-A16 prepared in PBS (MOI 1) for 1 h. Next, PBEC were washed by PBS from the apical side to remove unbound virus and incubated for 6, 24 and 48hpi respectively. As controls, air- and CS-exposed ALI-PBEC were mock infected with PBS.
For interferon treatment, ALI-PBEC from non-COPD donors (n = 4) were stimulated with 10 ng/ml recombinant human interferon β (IFN-β, R&D Systems, Minneapolis, USA) or 10 ng/ml IFN-λ1 (PeproTech, London, UK). These interferons were added in the basal medium directly after CS exposure, followed by RV-A16 infection (MOI 1) for 1 h and the cells were harvested at 24hpi.
For FCCP treatment, ALI-PBEC from non-COPD donors (n = 3) were treated with 10 μM FCCP for 6 h from apical side of the inserts and lysed for qPCR analysis. FCCP (carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone), served as a positive control based on its ability to cause a decrease in mitochondrial membrane potential (MMP).
For l-lactate treatment, ALI-PBEC from non-COPD donors (n = 3) were treated with 10 mM sodium l-lactate (Sigma-Aldrich) in the basal medium for 10 min. The concentration (10 mM) was chosen based on the levels of l-lactate measured in CS-exposed ALI-PBEC at 24hpi. Next, cells were apically infected with RV-A16 (MOI 1) at room temperature, followed by an apical wash after 1 h and incubation for 24hpi.
The knockdown of GDF15 was performed in the PBEC from non-COPD donors (n = 4) and after knockdown of GDF15, PBEC were cultured at ALI for 2 weeks and exposed to CS and RV-A16 with controls as described earlier and analyzed at 24hpi. GDF15 knockdown in PBEC was performed based on Alt-R CRISPR-Cas9 System [Integrated DNA Technologies (IDT), Coralville, IA, USA] using a combination of electroporation and lipofectamine delivery methods.
Detailed procedures of cell culture, cigarette smoke exposure, RV-A16 stock preparations, TCID50 measurements,
GDF15 knockdown, RNA isolation, qPCR, ELISA,
l-lactate measurements and lactate dehydrogenase (LDH) assay are provided in the Additional file
1: supplementary methods.
RNA sequencing and analysis
A dataset of 104 samples was generated by mRNA sequencing (RNA-Seq, polyA enriched) using the Illumina NovaSeq600 sequencer, with 15 million paired-end reads per sample which was performed at GenomeScan (Leiden, the Netherlands). All samples had a RIN (RNA integrity number) score of > 6 as assessed by bioanalyzer (GenomeScan) and passed the FASTQ and alignment quality analysis with OmiWCSoft (Qiagen, Venlo, The Netherlands). Detailed description of RNA isolation, sequencing and analysis along with Gene Set Enrichment Analysis (GSEA) [
14] and Ingenuity Pathway Analysis (IPA) are provided in Additional file
1: supplementary methods.
Cellular deconvolution
The estimation in the ratio of epithelial cellular composition, including ciliated, secretory, basal and rare cells, was obtained by cellular deconvolution analysis of bulk RNA-Seq data as previously reported [
10]. Briefly, based on minimized correlation and maximized distance between clusters in which genes with the most stable results across cohorts were selected and used to infer major cell type proportions, 400 genes from highly variable ones were selected and filtered according to the human Lung Cell Atlas v1.0 dataset [
11] using AutoGeneS software. The RNA-Seq data was subsequently normalized to counts per million (CPM), and highly variable (HV) genes (N = 5000) were selected. Cellular deconvolution was performed based on the RNA-Seq dataset using the CIBERSORT support vector regression (SVR) method [
12]. The relative proportion of cell types was compared between ALI-PBEC exposed to Air, CS, RV-A16 and CS combined with RV-A16 in COPD and non-COPD donors using 2-way ANOVA with Tukey’s or Bonferroni test.
Measurement of mitochondrial membrane potential
The mitochondrial membrane potential (MMP) was measured using the live detection probe JC-10 (Enzo Life Sciences, Farmingdale, NY, USA). After WCS exposure or FCCP (10 μM) treatment for 10–30 min, 2.5 µg/ml JC-10 was added to the apical side of the inserts for 30 min in the cell incubator and PBS was used to remove JC-10. The inserts were then cut out and placed on a Superfrost Plus object glass (ThermoFisher) covered with a coverslip. Live cells were imaged with a Leica TCS SP8 confocal microscope (Leica Microsystems) at 630× original magnification. For mitoTEMPO (Sigma-Aldrich) treatment, a concentration of 50 μM, was added to the basal medium 4 h prior to CS exposure or FCCP treatment. The ratio of positive-stained cells from three random areas of each insert membrane of each independent experiment were quantified by ImageJ.
In vivo RV-A16 challenge
The gene sets that were differentially expressed upon in vitro RV-A16 infection and/or CS exposure were analyzed using a previously published dataset [
13] from nasal epithelial cells derived of non-smokers, non-COPD donors (n = 16; only placebo treated group included) after in vivo RV-A16 challenge. We compared interferon response, complement, necroptosis and inflammatory gene sets from our study with those expressed in nasal epithelial cells collected on days 3, 6, 9 and 13 post RV-A16 challenge and at baseline (one day before RV-A16 challenge). RNA was isolated from nasal brushes during a previously published study [
13] containing predominantly nasal epithelial cells (~ 95%) and RNA sequencing (RNA-Seq) was performed in a similar platform as for the cultured ALI-PBEC. RNA-Seq of nasal brushes and its analysis was already published, which included the expression of interferon response gene set at day 3, 6, 9 and 14 compared to baseline (before RV-A16). The detailed information of in vivo RV-A16 challenge, method details and the inclusion and exclusion criteria of these donors can be found in a previous publication [
13]. At each time point, two nasal brushings obtained were pooled and centrifuged at 1000×
g (standard tabletop centrifuge) for 5 min at 4 °C. The pellet was dissolved in 1 ml of TRIzol (ThermoFisher) and stored at − 80 °C. After all samples were collected, they were thawed to room temperature, followed by addition of 200 µl of chloroform and inverted 10 times. The samples were centrifuged at 2000×
g for 10 min at 4 °C. The aqueous phase was used to isolate RNA using protocol 5.3 from RNA XS extraction kit (Macherey–Nagel, Düren, Germany). The quality and concentration of the samples were assessed by using a fragment analyzer (Advanced Analytical Technologies, Inc, Ankeny, Iowa).
Statistical analysis
Statistical analysis was performed in GraphPad PRISM 9.0 (GraphPad Software Inc., La Jolla, CA). Differences were explored by one-way or two-way ANOVA with Tukey’s test, paired or unpaired two-tailed, t-test. Data are shown as mean values ± SEM and differences at p values of < 0.05 were considered significant.
Discussion
In this study, we investigated mechanisms underlying the increased RV-A16 infection upon CS exposure of ALI-PBEC using a transcriptomics-based approach combined with functional studies. In line with various studies in other epithelial cell models, we confirm that whole CS exposure increased rhinovirus load in ALI-PBEC cultures. Importantly, we identified an impairment in expression of interferon response genes early after CS exposure, likely contributing to increased rhinovirus replication based on the ability of added interferons to reverse the effects of CS exposure on viral infection. Using functional studies, we additionally demonstrated that enhancement of GDF15 and l-lactate production after CS exposure contributes to increased viral replication. At a later phase, despite higher viral replication in CS-exposed ALI-PBEC, interferon response genes remained unaltered, while expression of epithelial interferons and inflammatory mediators were increased. Finally, analysis of data from a human experimental RV-A16 challenge study demonstrated that our results in vitro show overlap with similar pathways after in vivo RV-A16 exposure, including interferon response, complement, necroptosis and inflammation.
CS exposure of ALI-PBEC decreased expression of IFN response genes and PRRs implicated in viral recognition at 6hpi, which may have hampered antiviral defenses in the early stages of viral infection and thereby contributed to the observed increase in viral replication at 24 and 48hpi. One study demonstrated that early administration of IFN-β protects CS-exposed mice from lethal influenza virus infection [
17]. In ALI-PBEC, we showed that pre-treatment with either IFN-β or IFN-λ1 decreased RV-A16 replication after CS exposure, confirming their protective effect in an early phase. At 24hpi, RV increased expression of IFN response genes and CXCL10 production in ALI-PBEC. Another study showed that RV infection of bronchial epithelium leads to a virus titer-dependent expression of interferon-response genes such as
CXCL10 and
RSAD2 [
18]. Despite higher viral replication in CS-exposed epithelial cultures at 24hpi (which would be expected to trigger higher levels of CXCL10 and expression of interferon response genes), expression of these interferon response genes and CXCL10 protein levels were similar to infected cultures without CS exposure, indicative of a dampened interferon response. In contrast to these interferon response genes, expression of interferon genes in RV-A16 infected cells was increased by CS exposure at 24hpi. This effect likely relates to the higher viral load after CS exposure. Even though epithelial interferon production was increased at 24hpi, the viral replication was not decreased in CS-exposed ALI-PBEC at 48hpi, which may be related to the observation that interferon-response genes were not altered by CS exposure despite the higher viral load and higher interferon production. Rather, CS exposure increased inflammation and necroptosis and therefore increased interferon could potentially contribute to excessive lung inflammation and tissue damage at a later stage [
19]. Altogether, our studies indicate that there might be distinct effects of smoke on antiviral responses between early and late stages of RV infection.
We further identified CS-mediated mechanisms that contribute to rhinovirus infection and showed that CS exposure increased oxidative stress, which may contribute to mitochondrial damage [
20]. We hypothesize that this damage is responsible for the increase in the stress-induced cytokine/growth factor GDF15, that we found to directly contribute to impaired viral clearance. This hypothesis is supported by a previous study showing that a decrease in MMP impaired MAVS-mediated antiviral signaling in the mitochondria, leading to higher viral infection [
21]. We showed that CS exposure and FCCP (positive control for depolarization of MMP) both increased expression of GDF15 in ALI-PBEC. GDF15 levels are elevated in COPD patients [
22] and its epithelial expression is induced by CS [
23]. A link between GDF15 and viral replication was supported by a study in which
GDF15-overexpressing mice and GDF15-treated bronchial epithelial cells showed higher viral replication, which was attributed to an inhibition of IFN-λ1 [
15]. However, whether GDF15 contributes to the increased RV replication following CS exposure in ALI-PBEC was unknown. We demonstrated that CS increased both RV-A16 vRNA and GDF15 levels at 24hpi, and that IFN-λ1 levels were similar upon RVA-16 infection in both air and CS-exposed cells. Whereas we observed that partial knockdown of
GDF15 appeared to decrease the effect of CS exposure on RV replication, this was based on a limited number of donors and the differences did not reach statistical significance. Nevertheless, our data do not allow us to conclude that GDF15 mediates increased RVA-16 replication by decreasing IFN-λ1 expression, as IFN-λ1 expression is regulated by RV-A16 vRNA levels (which are increased upon smoke exposure). Future studies using e.g. CS exposure and polyI:C stimulation instead of RV-A16 infection to increase IFN-λ1 expression could shed more light on this, but are beyond the scope of the present study. Collectively, these findings suggest that GDF15, increased by CS directly by oxidative stress or indirectly by mitochondrial membrane depolarization, may contribute to the CS-induced increase in RV-A16 infection in ALI-PBEC.
Next, we investigated changes in transcriptional responses by CS at 24hpi and showed increased expression of a glycolysis-related gene set. A recent study showed that lactate derived from glycolysis was found to target the MAVS protein and thus inhibits IFN production and enhances viral replication in mice [
24]. In line, we showed that pretreatment of ALI-PBEC with sodium
l-lactate, which we found to be increased upon CS exposure, enhanced RV-A16 replication. Apart from glycolysis, we also demonstrated that oxidative phosphorylation was only enhanced by the combination of both CS and RVA-16 exposure. Therefore, we illustrate that CS exposure changes the glycolysis metabolic pathway, which further contributes to viral infection.
In our study, there was no significant difference in response of the exposures between COPD and non-COPD-derived cultures. We included cells derived from moderate COPD patients (Stage II) and not severe COPD patients (Stage III or IV) and have compared COPD and non-COPD using 8 donors in each group, which may have lacked sufficient power to reach statistical significance as some loss of phenotype in culture is to be expected. Some phenotype is however still present as deconvolution analysis showed a higher proportion of secretory cells at baseline in cultures from COPD donors compared to non-COPD controls. In COPD patients, immunohistochemistry staining of the large airways showed increased MUC5AC levels, a marker for goblet cells [
25]. Thereby, in line with our previous findings [
26], the ALI-PBEC cultures from COPD patients did indeed retain certain intrinsic features in culture. Furthermore, our analysis shows that CS exposure reduces the proportion of ciliated cells and cilia-associated gene expression, which may contribute to impaired mucociliary clearance [
27]. The deconvolution analyses are based on a short follow-up after CS exposure (up to 24 h), and therefore indicates predicted changes in cell proportion in ALI-PBEC at a later phase, rather than definitive alterations in the percentage of cell numbers. Therefore, it would furthermore be valuable to also investigate the effect of chronic CS exposure of the epithelium on viral infection for a closer relevance to COPD [
9,
28]. The results of this study do not provide evidence for a role of interactions among the pathways related to interferons, glycolysis and GDF15 in mediating the effect of CS on RV-A16 replication, but more studies are required for firm conclusion regarding such interactions. Furthermore, our model is limited because it only comprises (differentiated) epithelial cells, and the role of immune cells, bronchial smooth muscle cells (and other structural cells) was not investigated. RV-A16 can drive the immune response against viral replication by infecting bronchial epithelium or macrophages [
29] or bronchial smooth muscle cells [
30] directly. Use of co-culture models or ex vivo precision-cut lung slices, would be valuable to study cell interactions.
Overall, our study provides evidence for a functional involvement of early and late anti-viral responses, along with GDF15 and lactate production in CS-mediated increase in RV-A16 infection. These findings aid in unravelling the complexity related to this increased infectivity and may promote research into therapeutically targeting of these pathways for virus-induced COPD exacerbations.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.