Repeated nebulisation of non-viral CFTR gene therapy in patients with cystic fibrosis: a randomised, double-blind, placebo-controlled, phase 2b trial

Citation for published version: Alton, EWFW, Armstrong, DK, Ashby, D, Bayfield, KJ, Bilton, D, Bloomfield, EV, Boyd, AC, Brand, J, Buchan, R, Calcedo, R, Carvelli, P, Chan, M, Cheng, SH, Collie, DDS, Cunningham, S, Davidson, HE, Davies, G, Davies, JC, Davies, LA, Dewar, MH, Doherty, A, Donovan, J, Dwyer, NS, Elgmati, HI, Featherstone, RF, Gavino, J, Gea-sorli, S, Geddes, DM, Gibson, JSR, Gill, DR, Greening, AP, Griesenbach, U, Hansell, DM, Harman, K, Higgins, TE, Hodges, SL, Hyde, SC, Hyndman, L, Innes, JA, Jacob, J, Jones, N, Keogh, BF, Limberis, MP, Lloyd-evans, P, Maclean, AW, Manvell, MC, Mccormick, D, Mcgovern, M, Mclachlan, G, Meng, C, Montero, MA, Milligan, H, Moyce, LJ, Murray, GD, Nicholson, AG, Osadolor, T, Parra-leiton, J, Porteous, DJ, Pringle, IA, Punch, EK, Pytel, KM, Quittner, AL, Rivellini, G, Saunders, CJ, Scheule, RK, Sheard, S, Simmonds, NJ, Smith, K, Smith, SN, Soussi, N, Soussi, S, Spearing, EJ, Stevenson, BJ, Sumner-jones, SG, Turkkila, M, Ureta, RP, Waller, MD, Wasowicz, MY, Wilson, JM & Wolstenholme-hogg, P 2015, 'Repeated nebulisation of non-viral CFTR gene therapy in patients with cystic fibrosis: a randomised, double-blind, placebo-controlled, phase 2b trial', The Lancet Respiratory Medicine, vol. 3, no. 9, pp. 684-691. https://doi.org/10.1016/S2213-2600(15)00245-3


Introduction
Cystic fi brosis has been a target for gene therapy since the CFTR gene was cloned in 1989. 1 Lung disease is the main cause of morbidity and mortality in individuals with cystic fi brosis, with a median age at death of 29 years (95% CI 27-31). 2 Early expectations of a rapid breakthrough were based on supposed ease of access to the target respiratory epithelium via inhaled aerosols. These hopes were tempered by the subsequent realisation that the airways are well defended, in keeping with their predominant function as conducting passages, rather than absorptive surfaces.  Various vectors for delivery of the CFTR gene into respiratory epithelial cells have been assessed. Viral approaches, including adenoviruses, adeno-associated viruses, and retroviruses, have faltered because of ineffi cient transduction from the luminal surface and immune responses restricting the effi cacy of repeated application. 3 As such, research from the UK Cystic Fibrosis Gene Therapy Consortium has initially focused on non-viral vectors. Formulation and delivery of plasmid DNA-liposome complexes have been refi ned in a large series of preclinical studies, 4,5 and safety, 6,7 molecular effi cacy, and practical doses have been assessed in several phase 1 and 2a studies in patients with cystic fi brosis. 1,3 We did this study to assess the clinical effi cacy of the non-viral CFTR gene-liposome complex pGM169/ GL67A 8 after repeated delivery to the airways.

Study design and participants
We did this randomised, double-blind, placebocontrolled, phase 2b trial in two cystic fi brosis centres with patients recruited from 18 sites in the UK. Eligible participants had diagnosed cystic fi brosis, were aged 12 years or older, had a forced expiratory volume in 1 s (FEV 1 ) of 50-90% predicted, and had any combination of CFTR mutations.
The protocol was approved by the National Research Ethics Committee and the local Research Committees at the two dosing sites and the 16 other referral centres. Each patient, or a parent, provided written informed consent, and children provided assent.

Randomisation and masking
We randomly assigned patients (1:1), via a computer-based randomisation system, to receive nebulised pGM169/ GL67A or 0·9% saline (placebo). Randomisation was stratifi ed by % predicted FEV 1 (<70 vs ≥70%), age (<18 vs ≥18 years), inclusion in the mechanistic substudy, and dosing site (London or Edinburgh). Participants in the mechanistic substudy were randomly assigned (2:1) to receive nebulised pGM169/GL67A or placebo, and could participate as part of either a nasal or bronchoscopy group, or both. Participants and investigators were masked to treatment allocation, with the randomisation code known only by pharmacy staff at the two dosing sites.

Procedures
Patients received 5 mL of either 0·9% saline or pGM169/ GL67A complex nebulised through a Trudell AeroEclipse II device (Trudell Medical International, London, ON, Canada) at 28 day intervals (plus or minus 5 days) for 12 months. Each 5 mL dose of pGM169/GL67A contained 13·3 mg of plasmid DNA and 75 mg of the GL67A lipid mixture. Routine treatments were continued throughout the study, except for DNase, which was withheld for 24 h before and after dosing. In addition to the nebulised dose, patients in the nasal group of the mechanistic substudy received 2 mL of placebo or pGM169/GL67A divided between nasal cavities via a nasal spray device at the time of each lung dose. Patients in the bronchoscopy group followed the standard protocol, but also underwent a bronchoscopy under general anaesthesia before the fi rst dose and 28 days (plus or minus 5 days) after the fi nal dose.

Outcomes
The primary effi cacy endpoint was the relative change in % predicted FEV 1 , calculated from the mean of two baseline values (at screening and before dosing on day of the fi rst dose) to the mean of two values (2 and 4 weeks after last dose) at study completion. Secondary outcomes included additional measurements of lung function, CT scans, and Cystic Fibrosis Questionnaire-Revised (CFQ-R) scores. 9 Exploratory endpoints included exercise testing, activity monitoring, and sputum infl ammatory markers. Mechanistic endpoints were nasal or bronchial vector-specifi c DNA, mRNA, and electrophysiological assessment of CFTR function. We did extensive safety assessments.

Statistical analysis
The statistical analyses were prespecifi ed in a statistical analysis plan. With use of pilot data, we estimated the standard deviation of the relative change in % predicted FEV 1 in the target cystic fi brosis population to be 10% over 12 months. A total sample size of 120 assessable

Research in context
Evidence before this study We searched PubMed between June 1, 1992, and March 1, 2015, for studies published that included the terms "non-viral, gene therapy, cystic fi brosis" or "liposome, gene therapy, cystic fi brosis".

Added value of this study
We report the fi rst trial of non-viral CFTR gene therapy for patients with cystic fi brosis that is powered to detect clinically relevant pulmonary changes. Our study has progressed this fi eld of research from phase 1 and 2a studies showing changes in molecular surrogates of CFTR function, to a phase 2b setting assessing changes in lung function in patients with a broad range of CFTR mutations. Additionally, our study shows that monthly repeated application of non-viral gene therapy can be safely administered to the lungs over a 1 year period.

Implications of all the evidence
By providing the fi rst proof of concept that non-viral gene therapy can benefi cially aff ect lung function, follow-up studies can assess optimum dose, dosing interval, and patient stratifi cation at trial entry. Our fi ndings are likely to catalyse earlier translation of more effi cient vectors into fi rst-in-man trials. See Online for appendix patients would provide 90% power to detect a 6% diff erence between groups in the mean change from baseline at a two-sided 5% signifi cance level. This power calculation was conservative because covariate adjustment can be expected to increase statistical power. We did analyses in the per-protocol population (primary analysis), predefi ned as participants who received at least nine doses of pGM169/GL67A or placebo, and in the intention-to-treat population, who received at least one dose of pGM169/GL67A or placebo. We compared outcomes between groups with an ANCOVA model, with inclusion of the relevant baseline value, treatment allocation, and stratifi cation factors (baseline predicted FEV 1 , age, dosing site, inclusion in substudy). Results are reported as adjusted mean diff erences with corresponding 95% CIs. We assessed subgroup eff ects by including the relevant interaction term in the ANCOVA model. To allow results from diff erent endpoints to be plotted on a common scale, the estimated treatment eff ects were standardised and presented as multiples of the underlying SD. No adjustment was made to the p values to allow for multiplicity because the secondary endpoints were supportive and the corresponding p values were interpreted con servatively. We assessed bronchial and nasal biomarkers with a Mann-Whitney U test. A two-sided p value less than 0·05 was considered statistically signifi cant.
The trial was overseen by an independent Data Monitoring and Ethics Committee and a Trial Steering Committee. This trial is registered with ClinicalTrials. gov, number NCT01621867.

Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had fi nal responsibility for the decision to submit for publication. Figure 1 shows the trial profi le. Between June 12, 2012, and June 24, 2013, we randomly assigned 140 patients to receive placebo (n=62) or pGM169/GL67A (n=78), of whom 136 (97%) patients comprised the intention-to-treat population and 116 (83%) patients comprised the perprotocol population (fi gure 1). Reasons for discontinuation in the intention-to-treat population were similar between groups (appendix). Baseline characteristics were similar between the two groups (table 1). Unless indicated otherwise, all subsequent details relate to the per-protocol population.

Results
114 (98%) patients had paired pre-treatment and posttreatment measurements of % predicted FEV 1 . Of the two patients (both in the placebo group) who did not have paired measurements, one patient could not do the test because of a surgery-related pneumothorax and one withdrew because of time commitments and was unavailable for follow-up measurements. We recorded a signifi cant ANCOVA-adjusted treatment eff ect in the pGM169/GL67A group versus placebo at 12 months' follow-up (3·7%, 95% CI 0·1-7·3; p=0·046; fi gure 2) The relative changes within each of the individual groups were -4·0% (95% CI -6·6 to -1·4) in the placebo group and -0·4% (-2·8 to 2·1) in the pGM169/GL67A group (fi gure 2). Post-hoc analysis showed that 21 (18%) patients (n=6 in the placebo group and n=15 in the pGM169/GL67A group) had an improvement in % predicted FEV 1 of 5% or more of their individual baseline values. For comparison, the treatment eff ect in patients in the inten tion-to-treat  population who had spirometry measure ments both before dosing and within the protocol-defi ned window after their fi nal dose (n=56 in the placebo group and n=65 in the pGM169/GL67A group) was 3·6% (95% CI 0·2-7·0; p=0·039), with the 20 patients included in the intention-totreat, but not per-protocol, analysis, receiving a mean of 3·7 doses (SD 1·9). Figure 3 summarises changes in a range of secondary outcomes. The treatment eff ect was signifi cant for FVC (p=0·031; appendix) and CT gas trapping (p=0·048), but not for other measures of lung function, imaging, and quality of life (fi gure 3). We assessed whether a responder subgroup could be identifi ed; the appendix summarises the prespecifi ed subgroups. We noted no signifi cant diff erences in the primary outcome treatment eff ect with respect to sex, age, CFTR mutation (phe508del homozygous vs other), Pseudomonas colonisation, predominant smaller or larger airway disease on CT at presentation, concurrent drugs, or treatment-associated adverse events (appendix). Although some subgroups had larger treatment eff ects than others, these results were typically due to a greater decline in FEV 1 in the placebo group, rather than to any diff erence of eff ect in the pGM169/ GL67A group (appendix). Stratifi cation by baseline % predicted FEV 1 suggested a diff erence, albeit nonsignifi cant, in treatment eff ect between patients with more severe disease (FEV 1 49·6-69·2% predicted), who had a treatment eff ect of 6·4% (95% CI 0·8-12·1), and those with less severe disease (69·6-89·9% predicted), who had a treatment eff ect of 0·2% (-4·6 to 4·9; p interaction =0·065; appendix). In patients with more severe disease, post-trial and pre-trial changes in both the placebo group (-4·9%) and the pGM169/GL67A group (1·5%) contributed to the treatment eff ect. Secondary outcomes showed a similar trend favouring the more severe category (appendix).
Patients in both treatment groups received a median of three (IQR one to fi ve) courses of oral or intravenous antibiotics during the trial. Specifi cally, we assessed coadministered antibiotics during the critical analysis   period from dose 11 to the end of the trial. Numbers of patients receiving any additional antibiotics were 26 (48%) in the placebo group and 30 (51%) in the pGM169/ GL67A group (χ² p=0·774). Thus, the observed FEV 1 treatment eff ect was considered to be independent of concurrent antibiotic courses.
No clinically relevant pattern of changes could be distinguished in the exploratory outcomes of activity and exercise monitoring and serum and sputum infl ammatory markers (appendix). In the bronchoscopy group of the substudy, vector-specifi c DNA increased in 12 (86%) of 14 patients in the pGM169/GL67A group and was below the limit of quantifi cation in all (n=7) placebo samples (p=0·001; fi gure 4A); vector-specifi c mRNA was below the level of sensitivity in both groups (appendix). Changes in basal post-trial and pre-trial potential diff erence values did not diff er signifi cantly in either group (appendix). Figure 4B shows bronchial chloride responses using the mean of all interpretable tracings for each patient; a negative value indicates a change in the non-cystic fi brosis direction. Patients in the placebo group (n=7) had a median change (post-trial minus pretrial) of 3·1 mV (range 9·3 to -1·2) and those in the pGM169/GL67A group (n=10) had a change of -1·3 mV (4·0 to -5·8; p=0·032; fi gure 4B). Five (50%) of ten patients in the pGM169/GL67A group had values that were more negative than the largest response in the placebo group (fi gure 4). In the same analysis with only the most negative value recorded for each patient at any timepoint, patients in the placebo group had a median post-trial minus pre-trial change of 2·6 mV (range 9·3 to -1·2) and those in the pGM169/GL67A group had a change of -2·8 mV (4·0 to -16·8 mV; p=0·088; fi gure 4C). Six (60%) patients in the pGM169/GL67A group had values that were more negative than the largest response in the placebo group (fi gure 4). The appendix shows absolute bronchial potential diff erence values.
All patients had adverse events, with no signifi cant diff erence between groups for either total events or within the nine predefi ned adverse event categories (table 2). One patient in the placebo group and one patient in the pGM169/GL67A group discontinued study treatment because of adverse events (fatigue and increased respiratory symptoms and fl u-like symptoms, respectively). We recorded six serious adverse events, all in the pGM169/GL67A group (appendix). Neither the Data Monitoring and Ethics Committee nor the Trial Steering Committee regarded any serious adverse event as related to study drug; however, one event was considered to be possibly related to a trial procedure (bronchoscopy). We noted no clinically relevant changes in haematology, biochemistry, conversion of anti-CFTR T cells, anti-DNA antibodies, histology, or lipid staining (appendix) and no patients died during the study.

Discussion
We report the fi rst trial of non-viral based gene therapy for cystic fi brosis, powered to detect clinically relevant pulmonary changes. After monthly dosing for 1 year, we recorded evidence of a benefi cial eff ect of gene therapy versus placebo on FEV 1 . No eff ect of sex, age, or whether patients were homozygous for the most common F508del CFTR mutation could be detected. No clinically important adverse events attributable to treatment with pGM169/GL67A were reported.
Although these fi ndings are encouraging, they should be put into perspective. We noted a stabilisation of FEV 1 in the pGM169/GL67A group rather than an improvement. This stabilisation took place over a 1 year period and further work will be needed to see if this eff ect is maintained. The reduction in FEV 1 in the placebo group was within the range reported in some other prospective trials [10][11][12] and is consistent with a median survival of 29 years, but is greater than would be expected from registry data. 2 Three factors are likely to have infl uenced this diff erence. First, the requirement for clinical stability at trial entry meant that patients might have been at their optimum respiratory health at this stage. Second, the enthusiasm of patients to enter the trial, accompanied by a focus on self-care, might have resulted in short-term improve ments in lung function during the recruitment period. Both factors are likely to lead to a subsequent decline in lung function as patients regress to their mean values. Third, we included all available data, whether from stable patients or those with exacerbations, by contrast with registry data, which focuses on measurements obtained at annual review. Stabilisation of lung disease in itself is a worthwhile aim and we would caution against the bar being set too high for novel therapeutics in cystic fi brosis populations with an unselected range of mutations. The large response to ivacaftor in patients with class III mutations takes place in the context of correctly localised CFTR protein. By contrast, much smaller improvements in lung function were shown in the ivacaftor-lumacaftor trial for the most common mutation (phe508del) in which the CFTR protein is misfolded and mislocalised. 13 Placebo group (n=54) pGM169/GL67A group (n=62)   The response in our study was heterogeneous, with apparent responders and non-responders. The data suggest that an approximate doubling of treatment eff ect was achieved in patients with more severe disease stratifi ed by baseline FEV 1 , supported by trends in other clinically relevant secondary measures. A larger trial with a stratifi ed trial entry design, powered to assess subgroups, and that addresses the mechanisms of response heterogeneity, will be important to verify or refute these data. This diff erential response could relate to the dose deposited in the airways; in patients with lower baseline FEV 1 the relatively more obstructed smaller airways result in a larger proportion of the 5 mL dose being deposited in the larger airways. In pre-trial studies we assessed airway deposition in patients with cystic fi brosis with varying FEV 1 severity with technetium-99m labelled human serum albumin of similar droplet size (3-4 μm, using a diff erent nebuliser system) to the pGM169/GL67A formulation. Bronchial airway (generations 2-8) fractional deposition was 2·9% of delivered dose (standard error of the mean [SEM] 0·2; n=33) in patients with 70-90% predicted FEV 1 and roughly twice as great (6·0%, SEM 1·0; n=23) in those with 50-70% predicted FEV 1 . An additional contributory factor to this enhanced effi cacy might be the increased mitotic rate of more severely aff ected tissues, 14 which decreases the proportion of time that the nuclear membrane is intact, the membrane acting as a barrier to plasmid DNA entry to the nucleus.
We cannot rule out that the changes recorded in the present study are the result of a non-specifi c response to the pGM169/GL67A formulation. The placebo was 0·9% saline rather than a scrambled or CFTR-deleted plasmidliposome complex. We selected 0·9% saline partly on the basis of pragmatic fi nancial considerations, but mainly for ethical considerations, not wishing to expose patients with cystic fi brosis to fi rst-in-man repeated pulmonary dosing of an untested product that might direct the expression of an immunologically active peptide or novel non-coding RNA molecule with deleterious biological functions. Furthermore, we wanted to compare progression on therapy with the natural history of the disease. In terms of alternative explanations for the eff ects we noted, we know of no evidence that monthly nebulisation of 0·9% saline is deleterious to lung function, nor that liposome alone produces physiological improvements in either patients without, 15 or those with 16 cystic fi brosis. Delivery of non-CFTR encoding plasmid DNAs to the human airways has not been associated with a gain in CFTR chloride-channel function, nor improvement in any cystic fi brosis-related assay, 17,18 and plasmid DNA is generally associated with pro-infl ammatory, rather than non-specifi c, benefi cial eff ects. 19 We did not identify any pathophysiological changes in the airways, such as infl ammation or remodelling, nor any changes in bacterial species that might otherwise explain the outcomes. Nevertheless, we cannot exclude that DNA-liposome complexes augment host defences, stimulate mucus clearance, or enhance bacterial killing to an extent undetectable on semi-quantitative routine culture.
Results showing more robust changes in molecular CFTR surrogates would have been reassuring. Despite extensive optimisation of quantitative realtime-PCR assays, the pGM169-derived mRNA assay has poor sensitivity and is adversely aff ected by the inclusion of high levels of total RNA or modest concentrations of pGM169 plasmid DNA. In ovine studies we have shown that a 20 mL nebulised dose of pGM169/GL67A, four times that used in the present trial, is the lower threshold for reproducible detection of mRNA with this assay in airway tissue samples (unpublished). 6 Thus, our inability to detect pGM169-derived mRNA after delivery of 5 mL of pGM169/GL67A to the human airways, although disappointing, was not surprising. In human tissues, we have noted the low sensitivity of assays assessing vectorspecifi c mRNA from human samples in vivo, 16,20,21 and have noted the greater sensitivity of detection of electrophysiological changes, consistent with fi ndings in this study. 17,18,22 The ratio of area sampled to area dosed is small. Although we recorded signifi cant chloride secretory changes in the bronchial, but not the nasal, epithelium, we caution against placing undue weight on either observation. The size of the groups in the mechanistic substudy was limited by both the practicality of the procedures and the acceptability to patients of the additional invasive tests, leading to low statistical power for these measures. We would instead conclude that modest variable changes can be shown with currently available assays that remain insuffi ciently sensitive to detect changes in low levels of CFTR function when assessed in vivo in humans; further optimisation in these or other assays is needed.
Although we are encouraged by the fi rst demonstration of a signifi cant benefi cial eff ect in lung function compared with placebo associated with gene therapy in patients with cystic fi brosis, the mean diff erence was modest, only recorded in some individuals, and at the lower end of the range of results seen in clinical trials which result in changes in patient-related care. 23, 24 We did not formally assess infective exacerbations in view of the fairly small patient numbers in our study, but use of antibiotic courses as a surrogate identifi ed no obvious treatment advantage. The treatment eff ect is consistent with a clinically meaningful benefi t from the perspective of the European Medicine Agency; 25 however, further improvements in effi cacy and consistency of response to the current formulation, or its combination with CFTR potentiators, are needed before gene therapy is suitable for clinical practice. Furthermore, our fi ndings should encourage the rapid introduction of more potent gene transfer vectors into early phase trials, now that much of the groundwork has been established.