Human subjects were recruited as approved by the Institutional Review Board of Nationwide Children’s Hospital (IRB16-01020) and Emory University (STUDY00004965). All experiments were performed in accordance with relevant guidelines and regulations. Study subjects were provided written informed consent for procedures if of legal age, and children provided written informed assent and a parent or guardian of any child participant provided informed consent on their behalf. From clinic visits, clinical information was recorded into a REDCap database.

Figure 1 depicts the process followed in our study. Blood samples were taken during routine CF clinical examinations pre- and 3 months post-ETI therapy. Blood was immediately processed for plasma separation and frozen at −80°C until further processing. Samples were sent for analysis using the Olink^® Target 96 Inflammation Proteomics platform (Olink^® Bioscience, Boston), which measures 92 prespecified inflammatory proteins in as little as 1 μL of plasma. The Olink^® platform is based on the proximity extension assay (PEA) technology, where 92 oligonucleotide-labeled antibody probe pairs are allowed to bind to their respective target proteins, if present in the sample (32, 33). Olink^® uses a next-generation sequencing (NGS) approach, consisting of a PCR reporter sequence formed by a proximity-dependent DNA polymerization event. This was amplified and subsequently detected and quantified using real-time PCR. Internal and external controls are included in each sample for quality control of each step in the Olink^® protocol (two immunoreaction, extension, and amplification/detection controls). Samples that did not pass quality control were excluded from the analysis (n = 1). Data then underwent preprocessing normalization whereby values are set relative to a correction factor (Cq-value) to determine normalized protein expression (NPX) units generated on a log₂ scale.

Differential expression analysis (DEA) pairwise comparisons of NPX (matched to metadata) were performed using a Limma statistical method. The study design for DEA defined “group” as the primary factor, and no secondary factor (and blocking factor) was selected on ExpressAnalyst [https://www.expressanalyst.ca (34)]. Specific parameters consisted of a one-way ANOVA analysis with pairwise comparisons of non-CF (NCF) and pwCF before and after ETI matched to metadata of interest. The metadata collected and used for this analysis were as follows: study groups (groups: CF pre-ETI, CF post-ETI, NCF), responder to therapy (yes/no) based on ≥3% change in forced expiratory volume in 1 s (FEV₁) and 0.3 body mass index (BMI) or >10% FEV₁ alone (23), micro clearance of bacteria on clinical respiratory cultures (yes/no), prior CFTR modulator therapy (yes/no), results of the sweat chloride test (negative/indeterminate/positive), and genotype (heterozygous F508del, homozygous F508del, other). Differential expression values were used for different downstream cluster analyses to determine principal component analysis (PCA) variation profiles (https://www.metaboanalyst.ca), to create heatmaps of hierarchical clustering of differential NPX matched to metadata (ExpressAnalyst), and a supervised correlation heatmap of differential NPX (https://appyters.maayanlab.cloud/#/). Differential NPX data comparison (ExpressAnalyst) between clinical states was organized for visualization as volcano plots using SRplots [http://www.bioinformatics.com.cn/srplot (35)]. Differential NPX differences were considered significant with an adjusted p-value ≤0.05 and log₂ fold change of at least 1.2 or higher. Regulation of biological processes was determined using MetaCore enrichment analysis to identify pathways, process networks, and gene ontology (GO) processes impacted by ETI. Top biological processes were determined using GO pathway analysis on ExpressAnalyst and visually organized on SRplot. Clinical outcomes were compared using paired t-test statistics via GraphPad Prism v10.1. Significant differences in NPX were determined using a one-way ANOVA with an α cutoff of 0.05, and post-hoc corrections are indicated in the figure legends.

Variable importance projection scores for predictive modeling of inflammatory proteins that differentiated between clinical states were obtained with a multivariate receiver operating characteristic (ROC) analysis on MetaboAnalyst. We performed automated important feature identification and performance evaluation. ROC curves were generated with a Monte Carlo cross-validation (MCCV) with balanced subsampling. A support vector machine (SVM) multivariate algorithm was used as the classification method and univariate area under the receiver operating characteristic (AUROC) was used for average importance ranking.

To assess the effect of plasma from CF patients on the activation of transcription factors that control key physiologic pathways, STAT1-, STAT3-, STAT5-, and NF-κB-luciferase reporter cell lines were used (36, 37). These cell lines were cultured in DMEM with 10% FBS, 1% penicillin–streptomycin, and 0.2% Normocin. One day before the assay, cells were split and plated in a 96-well plate at 8,000 cells/well, in a volume of 50 µL per well. Following a 24-h incubation at 37°C, 30 µL of plasma from CF patients was added to each well, with four replicates per condition. Following a 30-min preincubation, each well was treated with either 20 µL of culture media as a control or stimulated with 20 µL of 1 ng/mL of IFN-γ for the STAT1-luciferase cells, 1 ng/mL of oncostatin M for the STAT3-luciferase cells, 10 ng/mL of prolactin for the STAT5-luciferase cells, and 10 ng/mL of TNF-α for the NF-κB-luciferase cells. Oncostatin M (300-10), IFN-γ (300-02), prolactin (100-07), and TNF-α (300-01A) were obtained from PeproTech US (Cranbury, NJ). Following a 6-h incubation at 37°C, luciferase activity was quantified using the Bright-Glo Luciferase Assay system (Promega Madison, USA) and a Luminoskan Ascent Luminometer (Labsystems, Vantaa, Finland).

Twenty pediatric pwCF and 20 healthy controls without CF were enrolled in the study. The basic demographics of the participants are listed in Table 1 including 1-year post-ETI data. Only 30% of pwCF had been on prior CFTR modulators. Pulmonary exacerbations were significantly decreased in the year after ETI initiation along with a significant sustained improvement at 1 year in lung function (FEV₁) and reduced bacterial colonization. BMI z scores were not significantly changed. Individual metadata at baseline, 3-month, and 1-year follow-up (CF only) of the participants are listed in Supplementary Table 1 .

After Olink^® quality control analysis (described in the Methods and Figure 1 ), a total of 74 of 92 proteins were detected in >75% of plasma samples. Global clustering protein profiles are shown in Figure 2 . Principal component analysis (PCA) was used to determine the proteome dynamics among the studied groups: pwCF before (CF pre), after (CF post) ETI therapy, and NCF controls. While the PCA showed distinct clustering of each group, the CF post-ETI group clustered between CF pre-ETI and NCF, with ~50% of CF post-ETI samples shifting toward an NCF profile ( Figure 2A ). A partial heatmap of overall protein expression profiles was generated using ward clustering ( Figure 2B , see Supplementary Figure 1 for a complete heatmap). The heatmap showed heterogeneity among protein expression in the population; however, a modest effect on several inflammatory proteins was observed.

To further assess protein expression changes, NPX data were used for protein DEA. We screened for proteins with significant changes in expression using both an adjusted p-value of 0.05 or less and a stringent log₂ fold change of at least 1.2 or higher. Unpaired analysis of NCF vs. CF pre-ETI resulted in 34 proteins with an adjusted p-value ≤0.05 and log₂ fold changes ranging between 0.24 and 1.7 ( Supplementary Table 2 ). Of these proteins, we identified four proteins with log₂ FC ≥1.2, namely, CCL20 (1.7 FC), MMP-10 (1.5 FC), EN-RAGE (1.3 FC), and AXIN1 (1.2 FC), which were upregulated compared to NCF ( Figure 3A ). Unpaired analysis of NCF vs. CF post-ETI resulted in 24 proteins with adjusted p-value ≤0.05 and log₂ FC range of 0.30 to 1.5 ( Supplementary Table 3 ). CCL20 was the only protein that remained upregulated after ETI when compared to NCF, at an FC of 1.5 ( Figure 3B ). In Figure 3C , it can be appreciated that although CCL20 decreased, the change was small and it remained upregulated. On the other hand, MMP-10 ( Figure 3D ) and EN-RAGE ( Figure 3E ) violin plots are comparable to NCF and distinguishable from CF pre-ETI. Meanwhile, the AXIN1 ( Figure 3F ) violin plot shows that CF post-ETI remains intermediate between CF pre-ETI and NCF. A paired analysis of CF pre-ETI vs. CF post-ETI identified three proteins ( Supplementary Table 4 ) whose expression was significantly changed with adjusted p-value ≤0.05 and log₂ fold changes ranging between 0.62 and 1.26. MMP10 with a 1.26 log₂ FC was downregulated post-ETI ( Figures 4A, B ). However, EN-RAGE (0.93 FC) and IL-17A (0.62 FC) expression decreased and their adjusted p-value was found to be statistically significant ( Figures 4C, D ). Combined, we identified modest changes in overall proteomics profiles post-ETI.

To identify which proteins impacted by ETI hold potential and therapeutic analysis, we performed a multivariate receiver operating characteristic (ROC) analysis in all three study groups. ROC ranks feature by average importance, which is a global measure of the ability of a test to discriminate whether a specific condition is present or not and find the best combination of parameters to predict group membership. ROC plots with area under the curve (AUC) values and confidence intervals for each model considered are summarized in Supplementary Figure 2 . In Figure 5 , we show the resulting top 15 proteins, ranked by average of importance (AI) between pwCF before and after ETI therapy; the closer the value is to 1, the more important the protein is considered. Most proteins were highly expressed (indicated in red squares) in pwCF before ETI compared to NCF with TNFRSF9 and CCL20 being the highest-ranked ( Figure 5A ) proteins in AI. Supporting our observation in Figures 3 , 4 , the proteins MMP-10, EN-RAGE, and IL-17A were also highly ranked in AI. Because TNFRSF9 and IL-17A had an AI higher than 0.8, it suggests that their FC can be of biological significance. A similar trend was observed when pwCF post-ETI were compared to NCF, where TNFRSF9 and CCL20 remained the highest-ranked proteins ( Figure 5B ). Additionally, all plotted proteins remained upregulated after ETI when compared to NCF. In contrast, when paired CF samples were compared, proteins CCL25 and TRANCE (TNFSF11), whose low protein expression is considered protective, were found to be upregulated after ETI compared to CF pre-ETI ( Figure 5C ). All the other plotted proteins, including the top AI-ranked proteins MMP-10, IL-17A, and EN-RAGE, were decreased post-ETI ( Figure 5C ). Furthermore, FGF21 is found to be downregulated post-ETI when compared to CF pre-ETI ( Figure 5C ). During fasting, FGF21 is released in the liver and was crucial for the regulation of glucose and lipid metabolism and energy homeostasis. A low expression of this protein is protective, in agreement with our results. On the other hand, we identified CCL11, ADA, FGF19, CCL25, TNFSF11, and EIF4EBP1 ( Supplementary Tables 2 - 4 ), whose low protein expression was considered protective, yet our data demonstrated that their expression remained upregulated not just in comparison to NCF but also CF pre-ETI. In conclusion, ROC analysis allowed us to support our initial observations and evaluate other proteins that fall out of the FC cutoff but can have biological significance and present potential topics for future research in CF.

CF pre- and CF post-ETI samples were further stratified by microbial clearance to evaluate whether bacterial eradication drives the reduction of proinflammatory proteins in people with CF post-ETI therapy. Among the 20 CF pre-ETI samples, 3 showed no evidence of infection, increasing to 6 samples post-ETI therapy with no bacterial infection. A closer analysis of these 6 matched samples revealed a total of 36 inflammatory proteins with a fold change of ≥1 post-ETI ( Supplementary Figures 3 , 4 ). However, these changes were not statistically significant (p ≤ 0.05). This aligns with the modest gene expression changes observed across the full set of 20 samples when comparing CF pre- and CF post-ETI groups. Therefore, we conclude that the decreased proinflammatory profile cannot be attributed solely to the increased bacterial eradication in some pwCF post-ETI therapy.

CF pre- and CF post-ETI groups were also stratified based on ΔppFEV1% ( Supplementary Figure 5 ). We observed a strong negative correlation between increased FEV1 and the expression of MMP-10 (r = −0.70) and IL-17A (r = −0.67). Additionally, the decreased expression of EN-RAGE also correlated with increased FEV1, though at a lower magnitude (r = −0.38). These findings further supported the potential of MMP-10 and IL-17A for biomarkers in CF.

To evaluate the relationships between the proteins identified during unpaired analysis in signaling pathways, we performed a Spearman rank correlation analysis. As mentioned above, unpaired analysis of NCF vs. CF pre-ETI identified 34 differentially expressed proteins and that of NCF vs. CF post-ETI revealed 24 differentially expressed proteins. Correlation analysis among inflammatory proteins revealed three clusters ( Supplementary Figure 6A ) displaying strong positive correlations. The first correlation cluster included members of the TNF superfamily: TRANCE, TRAIL, TNF, and IL-12B, a cytokine expressed by activated macrophages. The second cluster included ST1A1, AXIN1, and SIRT2, proteins previously associated with fibrosis (38–40). Also, in the second cluster were STAMBP, TNFSF14, IL-7, CD40, CD244, and LAP TGFβ₁, which are linked to impaired lung function and damage. TGFβ is considered a genetic modifier of CF and a severe lung pathology biomarker (41–45). The neutrophil chemokines CXCL1 and CXCL6 are included in the second cluster as well and are released by lung epithelial cells and found in high levels in CF (46, 47). Lastly, the third cluster included IL-17A, MMP-10, HGF, EN-RAGE, IL-6, and TGF-α, all proteins involved in the NF-κB signaling pathway.

Furthermore, we performed gene set enrichment analysis (GSEA) using the GO database to confirm the biological pathways that these proteins are involved in. Pathway analysis resulted in 380 hits for the biological process (BP) domain, 27 hits for molecular functions (MFs), and 35 hits for cellular components (CCs). Of these, 13 BPs, 3 MFs, and 1 CC had a significant enrichment score ( Supplementary Table 5 ). Biological processes that were found to be significantly enriched and are of interest to CF are involved in the regulation of intracellular transport, protein import into the nucleus, cell-to-cell signaling, regulation of lipid and cellular metabolic processes, and regulation of defense response, among others ( Supplementary Figure 6B , top). MFs significantly enriched and of relevance to CF include functions related to transition metal ion binding, zinc ion binding, and cation binding ( Supplementary Figure 6B , middle). Significantly enriched CCs include receptor complex components ( Supplementary Figure 6B , bottom).

For a closer look at the impact of ETI on the regulation of these pathways, we performed an enrichment analysis on MetaCore. Initially upregulated signaling pathways such as those involving the NF-κB pathway, IL-17 and Th17 cells, were downregulated post-ETI ( Figures 6A, B , top). ETI therapy resulted in the downregulation of process networks involving the Jak–STAT pathway, the immune response of Th17-derived cytokines, and T helper cell differentiation ( Figures 6A, B , middle). Lastly, there was a decrease in chemotaxis and migration ( Figures 6A, B , bottom) of different cell types such as monocytes. Interestingly, processes necessary for response and defense against bacteria and fungus were downregulated after ETI therapy ( Figure 6B , bottom).

To determine the physiologic effects of these changes in cytokines and other soluble factors, we performed a functional assay that quantitates the net effects of plasma proteins on the activation state of four key transcriptional pathways: NF-κB (reflecting the inflammatory pathway), STAT1 (reflecting the interferon pathway), STAT3 (reflecting the acute phase response pathway), and STAT5 (reflecting hematopoietic and immune cytokines). This integrated functional quantitation (IFQ) assay integrates effects from positive regulators (such as multiple cytokines activating a specific pathway and the presence of activating soluble receptors) and negative regulators (such as antagonizing cytokines or soluble binding proteins). While no consistent effect was seen on the STAT1, STAT3, or STAT5 pathways, there was a strong downward trend in the activating ability of CF plasma post-ETI on the activation of NF-κB ( Supplementary Figure 7A ). This is consistent with the changes detected in the analyses described above.