The reader will be able to
- •
Appreciate how the cystic fibrosis journey from reactive treatment to proactive correction of the fundamental molecular defect has been achieved, and how this journey can be modified to move to proactive treatment for child.
- •
See how in the 21st century we must move from pattern-recognition, umbrella diagnoses to diagnosing specific molecular conditions.
- •
Understand the different uses of mutation agnostic and mutation specific therapies, and the implications of loss of function and gain of function mutations for planning therapy.
- •
Describe how in vitro , cell-based systems are allowing the understanding of subclasses of mutations especially of ABCA3 and SpC , analogous to cystic fibrosis, and thus truly personalising therapy for child.
Abstract
Cystic fibrosis (CF) is a monogenic disorder cause by mutations in the CF Transmembrane Regulator ( CFTR ) gene. The prognosis of cystic fibrosis has been transformed by the discovery of highly effective modulator therapies (HEMT). Treatment has changed from reactive therapy dealing with complications of the disease to pro-active correction of the underlying molecular functional abnormality. This has come about by discovering the detailed biology of the different CF molecular sub-endotypes; the development of biomarkers to assess response even in mild disease or young children; the performance of definitive large randomised controlled trials in patients with a common mutation and the development of in vitro testing systems to test efficacy in those patients with rare CFTR mutations. As a result, CF is now an umbrella term, rather than a specific diagnostic label; we have moved from clinical phenotypes to molecular subendotypes. Children’s Interstitial Lung Diseases (chILDs) comprise more than 200 entities, and are a diverse group of diseases, for an increasing number of which an underlying gene mutation has been discovered. Many of these entities are umbrella terms, such as pulmonary alveolar proteinosis or hypersensitivity pneumonitis, for each of which there are multiple and very different endotypes. Even those chILDs for which a specific gene mutation has been discovered comprise, as with CF, different molecular subendotypes likely mandating different therapies. For most chILDs, current treatment is non-specific (corticosteroids, azithromycin, hydroxychloroquine). The variability of the different entities means that there is little evidence for the efficacy of any treatment. This review considers how some of the lessons of the success story of CF are being applied to chILD, thus opening the opportunities for truly personalised medicine in these conditions. Advances in knowledge in the molecular biology of surfactant protein C and Adenosine triphosphate binding cassette subfamily A member 3 (ABCA3), and the possibilities of discovering novel therapies by in vitro studies will especially be highlighted.
Introduction: chILD – Where have we come from, where are we now?
The prevalence of chILD is difficult to estimate but is probably around 1 in 100,000 children . Within the chILD umbrella are what are conventionally termed more than 200 entities. However, as will be shown, many of what we called entities are in fact themselves umbrella terms (below).
Two seminal papers sought to classify chILD in the 0-≤2 and 2–18 age groups based on lung biopsy findings . Among the groups proposed were conditions specific to infancy such as neuroendocrine cell hyperplasia of infancy (NEHI), and surfactant protein gene mutations. With the benefit of further knowledge, both are also umbrella, descriptive terms, not 21st century diagnoses. The initial classification manuscripts have moved the field forward significantly, but it has now become evident that they did not capture many entities . As more genetic causes of chILD were described, many of which were not biopsied , it has become clear that we need to step back from the increasing complexity of classifications ( Table 1 ) and look in more detail at chILD endotypes.
| Underlying Category | Comments | Treatment Approach |
|---|---|---|
| Genetic | Definite – mutation discovered Probable/possible – VUS present, positive family history | Understand sub-endotypes to plan treatment as far as possible |
| Environmental | Detailed history essential, includes vaping | Remove environmental stimulus +/- systemic steroids |
| Other known non-genetic causes | Examples include iatrogenic (medication reaction) and GM-CSF autoantibodies causing PAP | As far as possible, remove the underlying cause +/- systemic steroids; some specific therapies possible, e.g. GM-CSF by inhalation or subcutaneous injection |
| Yet unknown cause | Fruitful area for research | Only non-specific treatments possible |
The present position is that we have moved from pattern recognition (pulmonary alveolar proteinosis (PAP), for example) to specific molecular and genetic underlying diagnoses (in the case of PAP, mutations in the α- and β-chains of the granulocyte–macrophage colony stimulating factor receptor or circulating GM-CSF autoantibodies ). There are specific therapies for a few rare genetic chILDs: anti-IL6 and Janus associated kinase (JAK) inhibitors for STAT3 gain of function mutations , abatacept for chILD due to LRB4 mutations , JAK kinase inhibitors for interferonopathies such as SAVI . However, for most patients all we can offer is non-specific treatments such as systemic corticosteroids, azithromycin, and hydroxychloroquine . The purpose of this review is to see what can be learned from the therapeutic triumphs achieved in cystic fibrosis (CF) in the hope of transforming chILD treatment in the future.
The CF journey: From tragedy to triumph
Within the last decade, treatment of CF has been transformed from non-specific, reactive therapies to sub-endotype specific molecular treatments with an established, solid evidence base for efficacy . This has been achieved with highly effective modulator therapy (HEMT). This has come about by (a) obtaining a thorough knowledge of the molecular biology of the CF gene; (b) where feasible, carrying out definitive clinical trials in large numbers of patients; (c) the use of more sensitive tests such as multiple breath washout to demonstrate efficacy in those with mild disease; (d) the use of biomarkers, especially sweat chloride as a biomarker in very young patients and those with only mild disease; and (e) the development of in vitro test systems to determine efficacy. These last are most useful in patients with rare CF genotypes in whom randomised controlled trials will never be feasible.
The bedrock of progress in CF came from the realisation that ‘CF’ is an inadequate genetic diagnosis, and the individual molecular subtype needs to be defined ( Table 2 ) . Specifically, in subclasses wherein an abnormal CFTR protein is produced, attempts to correct function may be fruitful, whereas in classes I and VII where no CFTR is ever synthesised, such approaches are doomed to fail. Subsequently it has been realised that the classification is not as clearcut as initially thought, with a given abnormal protein possessing defects in more than one class . However, this extra layer of complexity did not preclude the discovery of transformative therapies.
| Molecular mechanism | Example | Treatment option | |
|---|---|---|---|
| Class I | No protein produced | G542X | Currently none; molecules over-riding PTCs being tested Gene or RNA therapy being explored |
| Class II | CFTR mis-folded in Golgi and ubiquinated, never reaches cell surface | DF508 | Triple therapy (Trikafta, Kaftrio) |
| Class III | Impaired gating | G551D | Ivacaftor |
| Class IV | Decreased conductance | R117H with 5 T in trans | Ivacaftor |
| Class V | Decreased quantities of protein at cell surface | Ala455Glu 3272–26A → G, 3849 + 10 kg C → T | Stabiliser/ Amplifier |
| Class VI | Decreased protein stability at cell surface | c.120del23 Q1412X | Stabiliser/ Amplifier |
| Class VII | No mRNA produced | 1717-1G->A Dele2,3 (21kB) | Currently none Gene or RNA therapy being explored |
The first such medication, ivacaftor was shown to be highly efficacious in Class III gating mutations ( G551D , subsequently extended to other, rarer Class III mutations ). It is salutary to reflect that had ivacaftor been given to all patients with CF, it would likely have been discarded as inactive. Precision medicine based on a detailed understanding of the pathophysiology of the disease is the key to transformative therapies. However, HEMT has dramatically changed the landscape for those eligible CF patients even with severe disease and promises to radically change prognosis when started immediately after diagnosis by newborn screening.
Biomarkers such as reduction in sweat chloride have been accepted by the FDA as evidence of efficacy in those who cannot perform standard efficacy measurements (for example, infants) and those with very mild disease, in whom conventional endpoints such as spirometry and even lung clearance index are insufficiently sensitive.
It should be noted that it is not just in CF that precision medicine has been shown to be vital. The early trials of prednisolone and inhaled corticosteroids in ‘asthma’ were negative and these treatments could have been discarded because they were tested in all comers with obstructive airway disease, and not purely in those with eosinophilic airway inflammation (reviewed in . Subsequently, the anti-IL-5 monoclonal mepolizumab was initially thought to be ineffective because it was tested in the wrong group of patients . Precision medicine is impossible without precision understanding of the disease.
chILD: What can we learn from CF?
CF has clearly taught us that the first step to precision medicine is to understand molecular pathology, and we are appreciating that within a specific chILD gene, there may be multiple mechanisms of disease caused by different molecular subtypes, just as with CFTR . At its crudest, gain of function and loss of function in the same gene may lead to very different patterns of disease ( , Table 3 ) but as with CF, within a gene with loss- or gain- of function mutations, there are different pathways to disease (below).
| Gene | Gain of function | Loss of function |
|---|---|---|
| TBX-4 | Late onset interstitial lung disease | Pulmonary hypertension, skeletal abnormalities |
| STAT3 | Interstitial lung disease | Hyper-IgE syndrome |
There are important differences between chILDs and CF. The prevalence of CF varies, but at least some mutations (e.g. DF508 ) are so common that definitive randomised controlled clinical trials can be performed. chILDs are so rare, at least two orders of magnitude rarer than adult ILD, that it is highly unlikely that large, randomised trials of treatment will ever take place . It could be argued that some conditions, such as Adenosine triphosphate binding cassette subfamily A member 3 ( ABCA3 ) mutations, are sufficiently common to allow adequately powered trials, but in fact this is very far from being a pathophysiologically homogeneous group (below). However, CF has taught us that compelling evidence can be accrued even in those with highly rare genotypes.
Conventional investigation of chILD very often ends with pattern-recognition, such as desquamative interstitial pneumonia on a lung biopsy. In some chILDs we have moved beyond pattern recognition to teasing out specific sub types ( Table 4 ). This has led to sub-type specific therapies, for example nebulised or subcutaneous GMCSF in patients with PAP due to anti-GMCSF auto-antibodies . Again, giving this treatment to all those with PAP would likely lead to it being discarded as useless.
| “Umbrella” description | Subtypes of the condition | Potential treatments |
|---|---|---|
| Neuroendocrine cell hyperplasia of infancy | Genetic: TTF-1, TERT, FOXP1 BAL defined subgroups, cause unknown Others | Supportive: oxygen and nutritional supplementation No pharmacological treatment of benefit |
| Pulmonary alveolar proteinosis | Anti-GM-CSF auto-antibodies (adult type, but seen in children)Disorders of surfactant protein metabolism (adults and children) ; SpB, SpC, ABCA3, TTF-1 GM-CSF receptor gene mutations (α- & β-chains) Metabolic: lysinuric protein intolerance, Niemann-Pick disease Otherer genetic: MARS , GATA-2 Associated with immune deficiency | Whole lung lavage Nebulised or subcutaneous GMCSF Whole lung lavage Stem cell transplantation Avoid useless whole lung lavage May be specific monoclonals May be specific therapies |
| Hypersensitivity pneumonitis | Allergic response to mouldy hay ( Micropolysporium Faeni, mouldy hay ), birds (pigeons, budgies, doves, others, including via parental clothes), fungi in air coolers (India)Environmental exposures (Korean humidifier disinfectant) Vaping Genetic – Telomere mutations, others? Unknown | Allergen avoidance Avoid exposure Avoid active and passive vaping |
For therapeutic purposes, it may be helpful to divide chILDs into those which are early onset and rapidly fatal and those presenting later with a more indolent course, as a way of illustrating contrasting approaches. This is not an absolute dichotomy – mild cases of alveolar capillary dysplasia with misaligned pulmonary veins (ACDMPV) presenting later are increasingly recognized, for example . The remainder of this manuscript considers diseases within these two groups as exemplars to stimulate further work.
Neonatal onset, rapidly fatal chILDs
These invariably genetic diseases are exemplified by virtually all babies with surfactant protein B ( SpB) mutations, and those with the spectrum of alveolar capillary dysplasia – congenital alveolar dysplasia. Because the prognosis is so poor, and death is such a clearcut endpoint, developing and deploying novel therapeutic approaches is ethically and practically less challenging than in diseases running a more indolent course. Strategies (as with CF) can be divided into mutation agnostic and mutation specific approaches.
Mutation agnostic therapies for neonatal onset, rapidly fatal chILDs
Currently, these therapies are far from clinical reality. Gene therapy is one possibility, although the technical challenges are formidable with a large gene such as ABCA3 . Other possibilities might include RNA and cell-based therapies, although as a note of caution, if disease is caused by a gain of function mutation, it may be that replacing normal protein may be an insufficient strategy, and the abnormal protein may have to be blocked or synthesis inhibited. A mutation agnostic approach was used in a murine conditional SpB – / – model in which SpB was expressed under doxycycline control . As expected, when doxycycline was withdrawn the mice died rapidly. The investigators chemically modified SpB mRNA by replacing 25 % of uridine and cytidine with 2-thiouridine and 5-methyl-cytidine respectively, to reduce firstly mRNA binding to pattern recognition receptors such as TLR-3 and -7, and secondly to reduce activation of the innate immune system. Modified mRNA was aerosolized twice weekly into the mice after doxycycline withdrawal which rescued the lethal phenotype. Currently however, this approach is not a clinical reality.
Mutation specific therapies for neonatal onset, rapidly fatal chILDs
Approaches to Sp gene mutations are discussed below. This section focuses on ACDMPV. This is a misnomer, since injection studies have shown clearly that the “misaligned” veins running in the bronchovascular bundle are in fact dilated bronchial veins supplied by abnormally persistent fetal anastomotic channels .
ACDMPV is most frequently the result of underlying mutations in the Forkhead Box F1 ( FOXF1 ) gene, encoding the FOXF1 transcription factor. The underlying molecular pathology includes heterozygous deletions within the gene but also in upstream regulatory elements, and single nucleotide and small insertion/deletion variants. A recent murine study importantly showed that the structural pulmonary phenotype of ACDMPV could be reversed by post-natal therapy . The model focused on the P49-Y53 region in FOXF1 . This region contains numerous point mutations linked to the disease. Specifically, the S52F FOXF1 mutation was studied. The authors generated a murine FOXf1WT/S52F using CRISPR/Cas9 gene editing (the homozygous knockout is embryo lethal). The heterozygous mice had poor weight gain, pulmonary hypoplasia and pulmonary hypertension, lung histology which was characteristic of human ACDMPV and some of the characteristic systemic abnormalities seen in human disease. They then showed that there was reduced STAT3 signalling in both murine and human ACDMPV lungs. Genome-wide ChIP-seq analysis showed a 72 % overlap between FOXF1 and STAT3 endothelial target genes. Finally, and importantly, they then injected intravenously nanoparticles containing STAT3 in the mice affected with ACDMPV. The nanoparticles were able to target endothelial cells and reverse the histological changes of the disease. The authors rightly caution that this approach may not be relevant to all genetic defects causing ACDMPV. However, the potential postnatal reversibility of the histological changes, which many previously thought were irreversible, and the fact that the nanoparticle approach has reached at least Phase 2 studies in some malignant diseases, makes this potentially an exciting approach.
Another approach is the development of small molecules which stabilise FOXF1 protein, challenging the view that transcription factors are very hard to target. TanFe ((4- (4- chlorophenyl)- 2- {[2- oxo- 2- (thiophen- 2- yl) ethyl] sulfanyl}- 6- (thiophen-2- yl) pyridine- 3- carbonitrile) reduces FOXF1 degradation by disrupting its interactions with HECTD1 . Initial high throughput screening was employed to identify candidate molecules, and after Western blot analysis of the twelve most promising compounds, TanFe was selected. Intraperitoneal injection into pregnant mice of TanFe increased FOXF1 in ACDMPV pups, increased angiogenesis and prevented mortality. TanFe also Induced angiogenesis in human vascular organoids derived from human pluripotent stem cells made from fibroblasts of an ACDMPV patient with a FOXF1 deletion. There was also efficacy in a murine acute lung injury model. Of course, the key experiment would be to see if postnatal therapy would reverse the disease, because most affected infants have new mutations and are diagnosed postnatally. However, if postnatal efficacy and safety can be established, there is the potential for wide application in other neonatal diseases characterised by impaired vasculogenesis and pulmonary hypoplasia, such as congenital diaphragmatic hernia and bronchopulmonary dysplasia.
The heterogeneity of potential mechanisms to correct ACDMPV was highlighted in another study using murine organoids. The SF2F FOXF1 mutation, found in human ACDMPV, stimulates WNT/β-catenin canonical signalling in Type 2 cells, leading to their proliferation and reduction of Type 1 cell differentiation. The effects were prevented by exogenous WNT5A . However, whether this would result in reversal of the ACDMPV lung histology was not reported, and more work is needed on the relevance of this pathway.
Later onset, more indolent chILDs
The exemplar diseases discussed here are mutations in Surfactant Protein C ( SpC ) and ABCA3 , where there is recent progress in understanding the molecular biology potentially leading to personalising treatment.
ABCA3 disease
ABCA3 is a complex gene which is a member of the same gene family as ABCA7 ( CFTR ) ; although superficially the proteins look similar, there are in fact many subtle differences. The disease is autosomal recessive and more than 150 disease associated mutations have been described . Functionally, ABCA3 is involved in SpB and SpC post-translational processing, and the gene expression is controlled by the transcription factor TTF-1. Initially, a pragmatic clinical classification was proposed : ‘severe’, characterised by neonatal onset and death or transplantation within a year for those homozygous for severe variants; and mild, whereby the presence of one or more mild variants were often associated with later onset and prolonged survival. Indeed. presentation later in life is often associated with a stable course for many years . This was a useful classification for prognosis but does not help with personalising treatment. However it does serve the purpose of identifying a bad prognosis group in whom DNA or RNA based therapies might be attempted.
Pathophysiology of ABCA3 disease
Early studies suggested that ABCA3 mis-sense mutations might lead to impaired trafficking or dysfunction of the protein, and another suggested there were Type 1 and 2 mutations, respectively characterised by abnormal localisation or normal localisation but reduced ATP hydrolysis . An in vitro study started from the premise that ABCA3 mutations could result in either functional defects of correctly localized ABCA3 or trafficking/folding defects whereby mutated ABCA3 is retained in the endoplasmic reticulum (ER) (analogous to CF Class II and classes III-IV mutations respectively). Misfolded ABCA3 retained in the ER is toxic and activates the unfolded protein response which ultimately leads to cell apoptosis . This is a well described mechanism in adult idiopathic pulmonary fibrosis . The investigators transfected human A549 cells with wild type (WT) ABCA3 , and the ABCA3 mutations R43L , R280C and L101P , all three being known, chILD causing mutations. As expected L101P was retained in the ER and was associated with markers of ER stress (BiP/Grp78 were upregulated, XBP1 splicing was induced, and there were increased markers of apoptosis (Annexin V/PI staining, reduced GSH, increased caspase 3 and 4). WT protein colocalized with lysosomal-associated membrane protein 3 (LAMP3) positive lamellar bodies and lamellar-body-like vesicles. R43L trafficked in the same way as WT, but R280C partially trafficked as WT, but was partially retained in the ER . Of note, CF mutations also commonly exhibit features of more than one class (above). There was no in vivo readout, but the study strongly suggests that ‘ ABCA3 mutation’ is, like ‘CF’, a description not a diagnosis.
A similar study, also without an in vivo readout, divided ABCA3 mutations into two classes : disrupted intracellular trafficking, termed a type I mutant, the CF equivalent being Class II, rather confusingly; and impaired ATPase‐mediated phospholipid transport into lamellar bodies (type II mutant, with no obvious equivalent CF class). They also used A549 cells, and compared WT with four ABCA3 variants ( c.418A > C;p.Asn140His , c.3609_3611delCTT ; p.Phe1203del, c.3784A > G;p.Ser1262Gly , and c.4195G > A; p.Val1399Met ). p.Asn140His, p.Ser1262Gly, and p.Val1399Met proteins were processed and trafficked normally and there were normal-looking lamellar body‐like vesicles, with, however, reduced ATPase activity. P.Phe1203del protein was processed normally, had reduced ATPase activity, and also the vesicles looked normal, but also colocalized with both ER and lysosomal markers, suggesting abnormal intracellular trafficking as well as phospholipid transport.
The effects of different mis-sense mutations were determined in another study, underscoring that even within the same class of mutation there may be multiple disease producing mechanisms . There were three different mechanisms: disruption of intracellular ABCA3 protein localization (c.643C > A, p.Q215K, c.2279 T > G and p.M760R); impairment of lipid transport of ABCA3 protein ( c.875A > T , p.E292V ; c.4164G > C, p.K1388N ); and unknown disease-causing mechanisms despite normal localization and lipid transport of the variant protein ( c.622C > T , p.R208W ; c.863G > A ; p.R288K and c.2891G > A and p.G964D ). These classes may not be homogeneous; it is certainly conceivable that there could be reasons other than protein misfolding for retention in the ER, or that retention could be within the cytoplasm. Finally, a four-classification structure has been proposed , to which the class ‘unknown’ should probably be added ( Table 5 ) and undoubtedly further classes will be suggested. Of course, these categorizations are ultimately only of value if they lead to improved patient outcomes, as so brilliantly demonstrated in CF.
