Leveraging Population Genomics for Individualized Correction of the Hallmarks of Alpha-1 Antitrypsin Deficiency

Abstract

Deep medicine is rapidly moving towards a high-definition approach for therapeutic management of the patient as an individual given the rapid progress of genome sequencing technologies and machine learning algorithms. While considered a monogenic disease, alpha-1 antitrypsin (AAT) deficiency (AATD) patients present with complex and variable phenotypes we refer to as the “hallmarks of AATD” that involve distinct molecular mechanisms in the liver, plasma and lung tissues, likely due to both coding and non-coding variation as well as genetic and environmental modifiers in different individuals. Herein, we briefly review the current therapeutic strategies for the management of AATD. To embrace genetic diversity in the management of AATD, we provide an overview of the disease phenotypes of AATD patients harboring different AAT variants. Linking genotypic diversity to phenotypic diversity illustrates the potential for sequence-specific regions of AAT protein fold design to play very different roles during nascent synthesis in the liver and/or function in post-liver plasma and lung environments. We illustrate how to manage diversity with recently developed machine learning (ML) approaches that bridge sequence-to-function-to-structure knowledge gaps based on the principle of spatial covariance (SCV). SCV relationships provide a deep understanding of the genotype to phenotype transformation initiated by AAT variation in the population to address the role of genetic and environmental modifiers in the individual. Embracing the complexity of AATD in the population is critical for risk management and therapeutic intervention to generate a high definition medicine approach for the patient.

Introduction

Recent advances in genome sequencing technology is transforming our understanding and application of genomics to monogenic rare disease, for example, developing molecular therapeutics that target specific genetic variants in cystic fibrosis (CF) and Niemann-Pick C1 (NPC1),^1-5 as well as to complex polygenic diseases^6,7 such as risk assessment for chronic obstructive pulmonary disease (COPD).⁸ In addition to genomics, the development of techniques to monitor other personal omics, for example, epigenomics,^5,9,10 proteomics,^11,12 metabolomics,¹³ microbiomics,¹⁴ environmental exposomics¹⁵ and human activity tracking,¹⁶ as well as rapid progress of machine learning (ML) analytical tools,^1,17,18 is already advancing deep medicine¹⁹ at a high-definition²⁰ and high-performance level.²¹ Here, the patient is viewed as an individual in a population rather than being treated as an average of the population as occurs in conventional medicine.

Alpha-1 antitrypsin deficiency (AATD) is a monogenic familiar disease that is driven by basic and clinical hallmarks (Figure 1), each of which is uniquely influenced by genetic diversity in the population. Each of these hallmarks illustrate the complexity of disease associated with each patient in terms of genetics, tissue pathology (liver, plasma, lung), time of onset, progression and the environment. AATD is caused by genetic variants (alleles) in the SERPINA1 gene which has been identified in all ethnic groups worldwide with a frequency of 1 in every 2500 whites of European descent.²² Thought of as a monogenetic disease, AATD patients present highly variable phenotypes at different time frames with unique mechanisms that involve both gain and loss of function in different tissues (Figure 1). Alpha-1 antitrypsin (AAT) is synthesized in the endoplasmic reticulum (ER) and secreted from the liver through the exocytic pathway in large quantities on a daily basis to maintain a plasma concentration at 1-2 g/L.^23,24 A well-established biological function of AAT is its antiprotease activity within the lung that prevents tissue degradation by neutrophil elastase (NE) and is currently the most established genome-wide association study (GWAS) modifier leading to COPD.²⁵ More recently, AAT has been shown to have anti-inflammatory and immunomodulatory functions independent of the antiprotease activity that may contribute to AATD phenotypes.^26-29

Approximately 95% of severe AATD cases are caused by the Z allele^22,30 where the glutamic acid residue is mutated to lysine residue at position 366 in the AAT polypeptide (sequence numbering including the signal polypeptide sequence). The Z variant leads to misfolding and polymerization of AAT in the ER of hepatocytes which can trigger hallmarks of liver disease phenotypes including chronic hepatitis, cirrhosis and hepatocellular carcinoma³¹ (Figure 1). Liver disease phenotypes are very heterogeneous in presentation with only about 10%-15% of infants with homozygous Z variant (PiZZ) developing clinically relevant liver disease,^32-34 although all homozygotes have some degree of accumulation of misfolded Z protein in the liver³⁵ upon aging and progression of the disease (Figure 1). Epigenetic mechanisms including DNA methylation¹⁰ as well as polymorphisms in ER mannosidase 1 gene (MAN1B1)³⁶ impacting glycosylation patterns, contribute to AATD liver phenotypes in subgroup cohorts³⁷ (Figure 1). There is also evidence that inherited traits influencing intracellular proteolysis mechanisms play a role in liver disease susceptibility. Due to the accumulation and degradation of Z-AAT in the ER of hepatocytes, a key hallmark of the disease is that only 10% to 15% of Z-AAT is secreted into the circulation (Figure 1). The loss of AAT function disrupts protease-antiprotease balance in the serum and lung, triggering emphysema and COPD³⁴ (Figure 1). Environmental factors, including cigarette smoke³⁸ and air pollution^39,40 contribute to the lung disease phenotype (Figure 1). Moreover, genetic modifiers including, for example, variants in matrix metalloproteinase 1 (MMP1),⁴¹ tumor necrosis factor,⁴² interleukin 10 (IL10),⁴³ iron regulatory binding protein 2 (IREB2) and cholinergic nicotine receptor alpha3 (CHRNA3)⁴⁴ impact the lung disease phenotypes in AATD (Figure 1). The heterogeneity and complexity of AATD presentation in the clinic indicates the importance of developing a high-definition, deep medicine²⁰ approach for each individual patient based on their genome sequence (Figure 1).

Herein, we briefly review the current therapeutic approaches for AATD. Not surprisingly, all efforts to date nearly exclusively focus on restoration of function of the dominant Z-variant polymerization. Thus, current approaches illustrate potential avenues for intervention for a single variant but fall short of the real problem– diversity in the population which makes it patient-unique. To address this problem, we highlight genotypic diversity of AAT variants and the impact of variation on the hallmarks of AATD (Figure 1) driving clinical presentation. Finally, we discuss the importance of understanding the genotype to phenotype transformation from the perspective of a new paradigm based on the concept of spatial covariance (SCV) in biology to individualize treatment of each patient from a genotype first perspective. In essence, we embrace the concept of “leave no patient behind,” a current goal of the Cystic Fibrosis Foundation,² as the operational paradigm for AATD treatment.

Current Generic Therapeutic Approaches for Z-Variant Alpha-1 Antitrypsin Deficiency

The vast majority of AATD therapeutic efforts target the dominant Z-variant allele. Here, the patient is treated as an average of the disease etiology of the Z-variant population based on either a reduced level of AAT in plasma and/or progression to endpoint liver/lung damage.

Enzyme Replacement

The current standard of care is based on the principle of enzyme replacement. Here, a quantitative increase in the serum pool of full-length AAT is mediated by intravenous infusion of plasma derived AAT (60mg/kg/wk), elevating transiently serum AAT to 50% of normal.⁴⁵ Treatment with 120 mg/kg/wk in a recent clinical trial has been shown to approximate a low-normal serum level of AAT with an associated increased antiprotease activity, decreased elastin degradation, and reduced airway inflammation.⁴⁶ A problem with this approach is that it fails to prevent the inherent accumulation and toxic effects of the expressed Z-variant in the patient.

Chemical and Biochemical Approaches

A number of small-molecule chemical and biochemical approaches are under development for AATD therapy with the goal of fixing the protein misfolding problem in the ER. These efforts largely focus on the aberrant polymerization events that leads to liver pathology and markedly reduces protein secretion into the plasma leading to reduced function in the lung.

One small-molecule chemical approach that addresses Z-variant aggregation in the liver makes use of the Food and Drug Administration (FDA)-approved drug carbamazepine (CBZ) that enhances autophagic pathways thereby reducing Z-AAT polymer intracellular accumulation in multiple cell and mouse models of the disease.⁴⁷ CBZ is currently in phase 2 clinical trials for individuals with severe liver disease (NCT01379469). ³⁰ Furthermore, the bile compound nor-ursodeoxycholic acid has been shown to clear > 70% of intrahepatic Z-AAT and reduced hepatocellular death through autophagy mechanism, suggesting a novel therapeutic approach for the treatment of liver disease in AATD.^48,49

An alternative approach to reduce aggregation involves the FDA-approved drug rapamycin, an mTOR inhibitor, that has been demonstrated to reduce the accumulation of hepatic polymerized Z-AAT and hepatocellular injury in a Z-AAT transgenic mice model.⁵⁰ Moreover, the generic proteostasis^51-53 drug trimethylamine N-oxide (TAMO) has been shown to protect Z-AAT protein from heat-induced polymerization in vitro, although it has no effect on Z-AAT protein secretion or protein re-folding in cell models.⁵⁴ In contrast, beside inhibiting Z-AAT polymerization and enhancing polymerized Z-AAT degradation, 4-phenylbutyric acid (PBA), as a proteostasis effector of unknown function, has been shown to promote Z-AAT protein secretion and function in both cell model and mouse models.⁵⁵ However, in 2 small human phase 2 trials in AATD patients, PBA had either no significant change on serum AAT levels, or led to a small increased serum AAT level in 11 PiZZ AATD patients.⁴⁵

While the above compounds are thought to directly impact proteostasis pathways, histone deacetylase inhibitors (HDACi), including suberoylanilide hydroxamic acid (SAHA), have been shown to increase Z-AAT protein secretion from epithelial cell lines to 50% of wild-type (WT) levels.⁵⁶ The exact mechanism remains unknown but could reflect epigenetic mechanisms (histone acetylation) leading to changes in the gene expression profile impacting protein folding programs and AAT stability in the ER, or post-translational mechanisms reflecting acetylation/deacetylation of AAT or membrane trafficking pathways facilitating AAT delivery in the cell.⁵⁶

An alternative chemical strategy is to prevent AAT polymerization using small molecules^57,58 and peptides.^59,60 AAT polymerization blocking molecules have been developed through virtual ligand screening against the hydrophobic cavity which is formed by s2A β-sheet, helix D and helix E in the AAT structure.^57,58 A number of reagents have been identified as potential binders that block Z-AAT polymerization in vitro and in the cell leading to improvement of liver pathology, although delivery of these reagents to intracellular compartments where the problem is initiated remains a major challenge.^30,58,61

In contrast to small-molecule chemical approaches impacting folding and stability of the Z variant, a biochemical approach involving the AAT-specific monoclonal antibody 4B12, which interacts with AAT protein by binding a conformation sensitive epitope involving residues Glu32, Glu39, His 43 and Leu 306, has been shown to block AAT polymerization in vitro, providing a potential approach to reduce the aggregation prone phenotype of Z-AAT.⁶² A single chain-variable-fragment intrabody has been shown to rescue Z-AAT function by both blocking intracellular Z-AAT polymerization and increasing functional Z-AAT protein secretion in cell and animal models, although the physiological mechanism remains a mystery given that intracellular Z-AAT pools in the ER are unlikely to be targeted by exogenous delivery of these reagents. ⁶³ However, these results provide evidence that monoclonal antibodies could be used as a therapeutic approach in treatment of AATD, possibly targeting Z-AAT that is secreted into plasma and in lung tissue to prevent further polymerization.^64-66

While all small-molecule chemical- or biochemical-based approaches are preliminary, each of these results raise the hope for small-molecule efforts for Z-variant AATD therapy by targeting mechanistically principles of fold design responsible for stability, monomer-polymer ratio and/or restoration of activity in the liver.⁴⁵

Genetic Therapy for Alpha-1 Antitrypsin Deficiency

An alternative strategy to the chemical- or biological-based interventions described above, are genetic therapy approaches. Here, the goal is to alter the genotype of the liver (or lung) to replace or convert the variant to a WT genotype in order to directly prevent aggregation disease in the liver and/or provide the lung with endogenously produced AAT to prevent destruction of the lung by NE and related protease activities, thereby reducing the impact of disease hallmarks (Figure 1).

In the past decade, a number of delivery vectors have been developed to administer WT AAT to the liver including adenovirus (AVV), retrovirus (RTV) and adeno-associated virus (AAV) vectors.^24,67 AVV gene transfer has been used to express human AAT protein in rat lung respiratory epithelium and hepatocytes.^68,69 RTV has been used for expression of human WT-AAT mice fibroblasts and dog hepatocytes.^70,71 Currently, AAV vectors are the most promising tool for gene delivery given their low toxicity, targeting specificity (liver, muscle, lung), and efficient long-term expression.⁷² Multiple serotypes have been used in preclinical AATD studies, whereas AAV1, AAV2 and AAVrh.10 have been used in clinical studies.⁷³ To date, AAV1 is considered the vector of choice given sustained expression of AAT, although in phase 2 clinical trials only 2% of the needed therapeutic level was sustained over 5 years, suggesting considerable ground for improvement.⁷⁴ A major drawback of these efforts is the inability to remove the toxic impact of the Z-variant allele in the patient.⁶⁷

To relieve the burden of accumulated variant protein-triggering liver damage, another strategy is to reduce the level of misfolded protein in the ER. Small interfering RNA (siRNA) and microRNA have been developed to efficiently silent hepatocyte human Z-AAT (hZ-AAT) protein expression and serum Z-AAT protein level in hZ-AAT mice through delivery by recombinant adeno-associated virus delivery vectors.⁷⁵ Moreover, antisense oligonucleotides targeting hAAT reduced Z variant plasma levels and reduced intracellular AAT protein accumulation in hZ-AAT transgene expression in mouse models. With long-term treatment, liver disease was reversed and a decrease in liver fibrosis was observed.⁷⁶ A similar result was observed in human PiZZ transgene expressing mice treated by AAV8 delivering AAT specific shRNA. shRNA decreased 90% of the AAT mutant protein.⁷⁷ Furthermore, siRNA targeting liver production of Z-AAT has been used in human clinical trials by 2 delivery vectors: ALN-AAT (NCT02503683) and ARC-AAT (NCT02363946) by Alnylam Pharmaceuticals and Arrowhead Research Corporation, respectively.⁷⁸ Phase 1/2 clinical trials using ALN-AAT have been conducted using homozygous Z-AAT allele patients since 2015. Clinical data shows a nearly 88.9% reduction of circulating Z-AAT.⁷⁷ In ARC-AAT clinical trials, the vector was tolerated and had sustained reduction of AAT (up to 90%) in a phase 1 trial.³⁰ These studies demonstrate that RNA interference represents a potential therapeutic approach for AATD liver disease but is limited by failing to fix the primary problem–the need for a sufficient level of functional AAT delivery to the plasma to protect against lung disease.

An alternative strategy to disease is to apply gene editing approaches such as CRISPR/Cas9 and related rapidly evolving technologies.⁷⁹ As a powerful tool for genome modification, gene editing systems have now been used to correct genome variants in multiple rare disease model systems.^80-83 Editing vectors transfected into Z-AAT transgenic PiZZ mice provides a proof-of-principle for the approach where hZ-AAT variant expression was reduced, circulating hATT protein levels were increased, and liver fibrosis and protein aggregation reduced.^84-86

Cell Therapy for Alpha-1 Antitrypsin Deficiency

While liver transplantation is the end-point strategy for treatment of liver disease, it has severe limitations given the need for donor tissue as well as the need for comprehensive immune suppression to prevent rejection. As an alternative approach to transplantation, AAT protein secretion has been observed by transplantation of hepatocytes from LacZ-transgenic ROSA26 mice into human Z-AAT transgenic mice. These results raised the possibility of hepatocyte transplantation as a therapeutic method for treatment of AATD in humans.⁸⁷ One possibility is that liver cells differentiated from stem cells from a normal individual found to express normal level of WT-AAT protein^88,89 could be used as a transplantation approach, although issues associated with immune rejection and toxic Z-variant load remain. Alternatively, human-induced pluripotent stem cells (hIPSCs) from the PiZZ patients could be differentiated into hepatocyte-like cells which express Z-AAT.^90,91 hIPSC-Z-AAT expressing cells show all the hallmarks of disease-disrupted mitochondrial structure, oncogenic protein AKR1B10 expression, upregulated inflammatory genes and induction of the unfolded protein response pathways.⁹² hIPSC Z-AAT can be corrected by gene editing using a combination of zinc finger nucleases and piggyBac with restored function and AAT secretion. When edited hIPSCs were transplanted into a mouse model of liver injury (Alb-uPA^+/+; Rag2^-/-; Il2rg^-/-) they were distributed throughout the liver leading to AAT secretion.⁹³ Moreover, edited human mesenchymal stem cells generated from patient stem cells have recently been shown to improve liver pathology as reflected in reduced inflammatory response and decreased fibrosis and apoptotic death compared to hZ-AAT-expressing hepatocytes in a mouse model.⁹⁴

Tackling Genotypic Diversity in Alpha-1 Antitrypsin Deficiency One Patient at a Time

While the Z-variant is the current focus of most therapeutic efforts, the complexity of AATD management is considerably amplified by the large number of variants now detected in the world-wide population through genotyping and genome sequencing efforts.^95-99 In the genome aggregation database (gnomAD, v2.1.1),⁹⁶ which includes the genome sequence information for 141,456 individuals, 601 variants in the SERPINA1 gene are reported. Among these reported variants, 277 are missense variants introducing amino acid residue changes in the polypeptide sequence with 116 being synonymous variants; 88 are in the intron region; 61 are in the untranslated region; 31 are splicing variants and the remaining 28 are deletion, truncation or frameshift. The top-3 most frequent variants are E400D (allele frequency [AF] = 27%), V237A (AF=22%) and R125H (AF=15.6%) which are historically referred to as M alleles (M3, M1 and M4 respectively) to indicate their WT-like activity. The S (AF=2.3%) and Z alleles (AF=1.1%) are fourth and fifth in allele frequency as missense variants, respectively.⁹⁵ The S allele (E288V) leads to a milder deficiency of AAT in plasma and the heterozygous PiSZ has a lower risk of lung disease than PiZZ individuals.¹⁰⁰ However, the clinical impact of the majority of rare variants is poorly understood or completely unknown. Thus, many alleles may be contributing in unknown ways to other features of disease etiology impacting the hallmarks of disease (Figure 1).

Besides the more common alleles listed above, we examined the distribution of 40 AAT missense variants found in the AATD population contributing to liver and/or lung phenotypes on the AAT structure, to illustrate the potential impact of genetic diversity AAT protein fold design (Table 1^101-147; Figure 2, upper panel). Here, patients carrying the variants illustrated in red have been reported to show both liver and lung phenotypes. Patients carrying orange variants have been reported with only the liver-associated phenotype while patients with yellow variants have been reported with only the lung-associated phenotype, whereas variant M382R, which is indicated in light blue, has a unique bleeding phenotype likely impacting the proteases in the coagulation cascade.^145,148-150 Variants found in the region of the reaction center loop adjacent to β-sheet A or in β-sheet A (Figure 2A, left panel, oval 1), such as E366K (Z), H358D and V357M, all have liver-associated phenotypes consistent with the known loop-sheet mechanisms leading to the polymerization of AAT in the liver that are predicted to impact AAT stability.^61,151,152 Furthermore, deletion of F76 (F76del) and S77F at the back of β-sheet A (Figure 2A, left panel, oval 1) also shows liver disease phenotypes, consistent with the interpretation that they may disrupt the stability of β-sheet A leading to an increase of the polymerization propensity of AAT.^146,153,154 Variants in β-sheets B and C (Figure 2A, left panel, oval 2) have more diverse phenotypes, for example, most of them are lung phenotype-associated variants, though some of variants also show liver-associated phenotypes. These variants are close to the ”gate” region which has been shown to allosterically mediate loop-sheet polymerization.^{147,152,155-158} Surprisingly, the variants localized to the bottom of the molecule (Figure 2A, left panel, oval 3) are exclusively associated with lung disease phenotype, indicating they do not cause aggregation in the liver but could affect secretion. Moreover, mapping these variants on AAT in complex with NE (Figure 2B) reveals that they are near the binding interface to NE, indicating these variants may even exclusively impact anti-protease (NE or other proteases) activity in the plasma or lung. It is possible that the many other variants have uncharacterized functions, for example, they may also contribute to ER-associated degradation promoting AAT deficiency, contribute to stability and/or translocation of AAT from the plasma to the lung, or modulate inflammation-related sensing in the lung.

Applying Deep Medicine to Individualizing Treatment in Alpha-1 Antitrypsin Deficiency

We now appreciate that the genome encodes key, evolutionary conserved information that reflects protein fold design principles that are largely unknown, yet define the functional requirements driving natural selection in response to a changing environment.¹

To appreciate how the genome encodes these protein fold design principles, the Balch laboratory has developed variation spatial profiling (VSP) based on the principle of SCV¹ (Figure 3A). VSP is a Gaussian process regression (GPR)-based machine learning (GPR-ML)¹ approach that re-describes central dogma in the context of SCV matrices that link sequence-to-function-to-structure.¹ Here, SCV relationships are defined by a sparse collection of variants found in the extant human population for any protein (such as the 40 variants found in AAT) (Figure 2). We refer to these known variants as trusted or fiduciary reporters, as they record the evolving genomic rules driving protein fold design. Using these variants to train the hidden layers using GPR-ML (Figure 3A), we generate sequence-to-function descriptions of the entire protein fold, that we refer to as “phenotype landscapes”¹ (Figure 3A and Figure 3B). Phenotype landscapes can be generated by using the sparse collection of variants from any protein to build SCV relationships based on their sequence position in the genome relative to functional features. For example, we have pioneered application of SCV using the rare disease protein cystic fibrosis transmembrane conductance regulator (CFTR). CFTR, like AAT, traffics through the secretory pathway. At the cell surface, instead of being secreted like AAT, it functions as a Cl^- channel function (Figure 3A and Figure 3B)¹⁵⁸ When defective in disease in response to over 600 variants in the population,^3,159 patients present with markedly diverse clinical phenotypes, even for the same prominent F508del allele,^2,3,160 These known relationships are then used to predict the function of all uncharacterized residues in the protein sequence design (Figure 3A and Figure 3B) that is displayed as a phenotype landscape and which can be mapped to protein structure at atomic resolution (detailed methods are explained in Wang et al¹) (Figures Figure 3A and Figure 3C). Phenotype landscapes are very robust in design, allowing the construction of phenotype landscapes that measure and predict the protein sequence response on a residue-by-residue basis to a broad spectrum of basic and clinical functional features, and in response to the changing environment that drives onset and progression of human disease.^1,161

As an example of the utility of SCV relationships in understanding therapeutic development and with the goal of understanding the impact of genetic diversity on management of AAT in the clinic, we generated phenotype landscapes based on 63 CFTR variants¹ to reveal the function of all residues across the entire CFTR sequence, illustrating that some variants effect ER export, while others only impact chloride channel function at the cell surface (Figure 3B,left panel). Whereas the most common variant, F508del, prevents ER export, the rare CFTR variant G551D was the basis for the first FDA-approved personalized therapy for CF–the drug ivacaftor (Figure 3B and Figure 3C).^159,162 G551D is a gating mutation that does not affect trafficking to the cell surface, as observed for some AAT variants that are efficiently secreted yet lack function (Figure 2). We have shown, using SCV relationships, that 63% of CFTR residues found in the disease population could be corrected by ivacaftor (Figure 3C).¹ More recently, in combination with other compounds targeting different features of the CFTR folding trajectory based on our current knowledge of the genotype impact on CFTR function,^1,3,159 a new triple combination of drugs improved the response of even the most severe F508del variant,^160,163-165 indicating that nearly all variants may be differentially accessible to correction in the clinical setting based on a complete understanding of SCV relationships contributing to disease. A second example are the rare variants found in the NPC1 protein (600 patients world-wide) responsible for NPC1 disease which maintains cholesterol homeostasis in all cell types.^166,167 Over 300 variants are now known to contribute to the differential onset and progression of cholesterol mismanagement leading to early onset neurodegenerative disease.⁴ Application of GRP-ML to NPC1 disease, using around 50 variants triggering disease, reveals that both proteostasis⁴ and epigenetic⁵ approaches can be used to correct cholesterol deficiencies of many but not all variants, revealing the power of SCV relationships in addressing the individualized nature of human disease where each patient in essence is unique and therefore requires a comprehensive understanding of disease phenotype.

Managing Alpha-1 Antitrypsin from a Spatial Covariance Perspective

AAT is a metastable protein and, like the CFTR transmembrane channel that goes through rapid open and closed cycles for chloride conduction, is a remarkable example of a conformation that has evolved to be sufficiently stable to support its synthesis in the ER of the liver through the exocytic pathway and secretion to plasma,¹⁵¹ while still maintaining its flexibility to perform in post-liver environments with its structural acrobatics that prevent the function of NE (or other proteases) in the lung in response to a large conformational change of molecule.^152,161,168 This dynamic conformational change is also, from a genetic diversity perspective, the Achilles heel of the disease.

Given the above, it is now apparent that different AAT variants will need to be understood in the context of their basic fold design principles from both basic and clinical sequence-to-function perspectives that can be captured by profiling SCV relationships.⁴ For example, the plot of variants on the known structure of native and NE-bound AAT (Figure 2) already suggests that different variants and their projected SCV relationships defining the AAT fold will have functions that are likely independent of one another, not only given the physical spatial differences associated with their sequence position, but the contribution of a given variant for a particular known and/or unknown function in the complex environments of host (i.e., liver versus plasma versus lung). For example, the liver is involved in nascent synthesis and management of fold monomer solubility at high concentrations in the ER for efficient secretion that is differentially challenged by different variants, whereas the post-liver function in plasma and/or lung environments is critical for inhibitor activity driving disease presentation and progression in the clinic (Figure 2).

Thus, by selectively applying knowledge of genetic diversity found in the extant AATD population through building SCV relationships based on GPR-ML, it may become possible to tailor therapeutic development and management to specific alleles that are unique to an individual. Such a “genotype first” approach will allow us to specifically and differentially, mitigate the hallmarks of AATD in an individualized, deep medicine approach (Figure 1).

Summary and Perspective

What is the best approach to manage AATD? Current approaches are illustrated in the context of hallmarks of AATD (Figure 1) and genetic diversity in the population with the genotype of the individual we posit being the ultimate gauge determining the basis for therapeutic management (Figure 4).^19-21

Genetic approaches are currently at the forefront of personalized approaches (Figure 4). For example, silencing of variant expression through RNA interference technology is the farthest along in clinical trials for Z-variant liver disease with a strong likelihood of success in the near term.⁷⁸ The limitations of this focused strategy include the lack of a means to restore expression of WT ATT. In contrast, gene editing, now possible through multiple approaches including CRISPR-Cas9 improved probes, still remains at its infancy for human use, particularly in terms of targeting efficiency and off-target effects that could exacerbate disease and/or trigger other unexpected pathologies. A key limitation of both of these technologies is not only expense but deciding when to initiate treatment, particularly as a pre-emptive strike before tissue damage in either the liver or lung is beyond repair. As an alternative to handling genetic diversity, patient-specific stem and iPSC cell supplementation or replacement strategies provide a very different and potentially powerful approach, but will certainly require considerable advances in current technologies given our poor understanding of their stability in the patient and malignancy potential, the nature of the tissue niche to which they need to be targeted and maintained, and proper management of the level of expression. Moreover, this approach lacks the ability to remove the toxic threat of the AAT variant driving disease, a concern inherent in current enzyme replacement approaches as well as the genetic therapies discussed above.

Small-molecule chemical/biochemical approaches also hold great promise to manage genetic diversity in the population that is unique for the individual (Figure 4), particularly if screened for effectiveness in the context of genetic diversity of the AATD population.^1,4,5 However, they are generally limited by their associated dosing and safety issues in complex environments of the human, often reflecting an inability to target the drug to the site of interest. Ivacaftor treatment for G551D and the predicted 63% of CFTR residues that are likely to respond to the drug based on GRP-ML (Figure 3), as well as the more recent triple drug combination for management of CF,^2,160,165 are pioneering examples of a personalized genotype specific strategy that targets different variants through their impact on the protein fold to minimize off-target effects. The Balch laboratory has proposed^1,4,5 that by understanding how human evolution works to manage diversity leading to fitness in the population in response to function of the protein fold,^1,4,5 we will be able to “pre-train” a high throughput screening drug development program based on genetic diversity to “pre-tune” the therapeutic approach in advance of clinical efforts.

Given the above, the first challenge we posit will be to lay the groundwork for precision disease management in the context of genetic diversity in AATD by understanding the impact of given variants in the context of the operation of the full-length fold, as reported by detailed natural history and clinical perspectives of the AATD patient population sequenced using whole-genome sequencing technologies (i.e., AllofUS).¹⁶⁹ Additional detailed biochemical/functional characterization of AAT variants are essential to understand the comprehensive genotype to phenotype landscape of variant AAT in response to different tissue environments as we have shown for CFTR¹ and NPC1^4,5 from genetic, epigenetic and therapeutic perspectives. We anticipate a major component of AATD treatment in the future will be a genotype first approach revealed by whole-genome sequencing efforts that help us to understand not only an individual’s AAT genotype but integrate that genotype with the many divergent basic and clinical features of disease reflected in the genetic diversity of the rest of the genome that make each one of us unique. These data will allow us to computationally formulate a common multi-dimensional framework to prioritize disease features that need to be corrected as well as their short- and long-term impact on the patient.

Put in another way, success will come from not only a deep understanding of the phenotype transformation driving disease in response to a patient’s unique AAT genotype, but will likely be influenced by multiple non-AAT alleles that contribute to the modifier environment that is, like AAT specific alleles, also unique to each individual harboring disease. Characterizing this complex interplay of patient specific genomics with their unique phenomics (Figure 4) in the context of environmental factors such as smoking, air quality and life-style will contribute to a more globalized understanding of each patient as an individual for personalized treatment. Digitizing AATD in the context of multi-dimensional layers of -omics matrices as performed using GRP-ML and rapidly advancing deep learning technologies^18,20 we predict will bring a new era of personalized medicine to AATD patient health and disease management. These are not insurmountable goals given the rapid advances now beginning to drive the future of deep medicine,¹⁹one that embraces the need for active patient involvement in all aspects of therapeutic development, as the SCV strategy is about “this is me.”¹

Abbreviations

alpha-1 antitrypsin, AAT; alpha-1 antitrypsin deficiency, AATD; machine learning, ML; spatial covariance, SCV; cystic fibrosis, CF; Neimann-Pick C1, NCP1; chronic obstructive pulmonary disease, COPD; endoplasmic reticulum, ER; neutrophil elastase, NE; genome-wide association study, GWAS; carbamazepine, CBZ; trimethylamine N-oxide, TAMO; 4-phenylbutric acid, PBA; histone deacetylase inhibitor, HDACi; suberoylanilide hydroxamic acid, SAHA; wild-type, WT; adenovirus, AVV; retrovirus, RTV; adeno-associated virus, AAV; small interfering RNA, siRNA; human Z-AAT, hZ-ATT; human-induced pluripotent stems cells, hIPSCs; human mesenchymal stem cells, hMSCs; allele frequency, AF; variation spatial profiling, VSP; Gaussian process regression, GPR; cystic fibrosis transmembrane conductance regulator, CFTR

Funding Statement

WEB is supported by a grant from the National Institutes of Heath (HL141810). JT is supported by the Alpha-1 Foundation. The meeting leading to this perspective was generously sponsored by the Alpha-1 Foundation.