Determining biomarkers for evaluation and diagnosis of hereditary angioedema

Abstract Rationale Kallikrein‐bradykinin‐forming cascade is known to cause hereditary angioedema (HAE) acute angioedema (AE) attacks. Further research of HAE attacks is needed to explain disease heterogeneity, predict treatment response and identify biomarkers for monitoring HAE attacks. Differential expression of the microvascular endothelial cell‐surface receptors for example, g‐C1qR, cytokeratin‐1, and plasminogen‐activator‐urokinase‐receptor (PLAUR) were hypothesized as biomarkers of AE attacks. Method To understand HAE attacks, the differentially expressed genes (DEGs) in RNAseq and mi‐RNAseq data of total RNA extracted from skin biopsies of lesional versus non‐lesional skin collected during and between attacks in Type‐1 HAE patients (n = 11; F:M = 8:3) were compared. To understand the HAE variants, DEGs in skin biopsies from HAE with normal C1 inhibitor (n = 5, F:M = 5:0), and non‐HAE (n = 7; F:M = 3:4) patients were compared. Gene‐set enrichment analyses and regulator effects analysis of these DEGs identified biological pathways in HAE attacks and their regulators. Results PLAUR gene, encoding urokinase‐type plasminogen activator (u‐PAR), was constitutively over‐expressed in HAE‐Type‐1 versus non‐HAE controls suggestive of overactive u‐PAR‐mediated signaling via binding to Factor‐XII. Baseline PLAUR expression was associated with severe AE (p = 0.05). The 18 significant DEGs investigated between baseline and AE attack samples in Type1‐HAE were enriched in beta1/beta3‐integrin cell surface interactions and IL‐6‐mediated signaling. Regulator effects analysis suggests a role for IL‐1b in HAE flares. AKT2, the mRNA regulated by the differentially‐expressed miR‐184A, was also associated with HAE attacks. Conclusion Angiopoetin‐activated β1‐integrin signaling pathways causing endothelial destabilization, and avid binding of factor XII to u‐PAR are possible novel mechanisms for progression of the endothelial kinin‐bradykinin‐forming cascade in HAE attacks.


| BACKGROUND
The kallikrein bradykinin-forming cascade is recognized as the primary mechanistic pathway for hereditary angioedema (HAE) swelling attacks. Although more effective HAE therapies have evolved over the past 15 years, further research of disease mechanism(s) is needed to explain disease heterogeneity, predict treatment response and identify biomarkers for monitoring disease during and between attacks. Clinically, characterization of HAE swelling attacks is difficult to predict because of their heterogeneity within and between patients with respect to frequency, location, severity and triggers such as psychological and physical stimuli that induce an episode. 1 Thus, there are current knowledge-gaps regarding biomarkers that can predict the onset and severity and response to treatment of HAE attacks.
The pathogenesis of HAE involves complex interaction of the complement, contact and fibrinolytic pathways. The primary mutations for HAE Type I and II involve the serine protease inhibitor G1 but more recently a number of genetic factors including factor XII (F12), plasminogen (PLG) and mutations in angiopoietin 1 have been found to result in increased bradykinin 2 receptor mediated signaling leading to increased bradykinin production. Additionally, ANGPT1 gene is known to perturb the cytoskeletal assembly of vascular endothelial cells. 2 Because of the possible involvement of the endothelial cell microenvironment in the pathogenesis of HAE flares, it was hypothesized that differential expression of the microvascular endothelial cell-surface receptors such as g-C1qR, cytokeratin-1, and plasminogen activator urokinase receptor (PLAUR) could be novel biomarkers for disease activity. These molecules exist in the endothelial cell membrane as the biomolecular complexes, gC1qR-cytokeratin 1 and cytokeratin 1 -urokinase-type plasminogen activator (u-PAR), which primarily bind high molecular weight kininogen-prekalikrein complex and F12, respectively. 3 This study aimed to determine the expression profiles of endothelial cell surface receptors, (e.g., plasminogen activation receptor -uPAR) from skin biopsies of patients with HAE Type 1 and HAE with normal C1 inhibitor (HAEnCI) between and during swelling attacks. In addition, investigation of the upstream regulators and molecular networks for such differentially expressed genes (DEGs) was performed to understand mechanisms of acute angioedema attacks based on canonical pathways from existing medical knowledge. Finally, using the regulator effects analysis, we also predicted novel pathways from predicted upstream regulators to downstream molecular/biological functions based on their common association with the DEGs.

| Demographics of study subjects
To understand the mechanisms of HAE Type 1 angioedema attacks (referred to as flares in this manuscript), skin biopsies from consented HAE Type 1 subjects (n = 11; F:M = 8:3) were collected during flares and during quiescent (asymptomatic) periods that were separated by at least 7 days after completion of the flare.
These biopsies were collected from edematous skin at the site of the HAE flare and from an adjacent area of skin after complete resolution of the attack. Skin biopsies during and in between swelling episodes were also obtained from HAE patients with normal complement (n = 6, F:M = 5:0). The HAE normal complement subjects all endorsed a family history for angioedema and were tested negative for the F12 mutation. Screening for other known mutations was not performed. A single skin biopsy was obtained from non-HAE patients (n = 7; F:M = 3:4) for comparison to HAE Type I samples.

| Inclusion criteria
(1) Ages 18-75 years, (2) confirmed HAE Type 1 (low C1INH), or HAEnCI for the comparison group, (3) patients using on-demand therapy only, and (4) ability to provide informed consent. Non-HAE subjects without other allergic disorders were enrolled as the control group.

| Collection of skin biopsy
See Appendix MethodsS1.

| Collection of blood samples and RNA purification
Five ml of whole blood was collected in RNA stabilizing blood tubes and stored at −80°C until further processing (see Appendix MethodsS1). Total RNA with miRNA were extracted from the blood sample using the PAXgene blood miRNA procedure as described in the vendor's manual (Qiagen).

| RNA-seq for differential gene expression profiling in skin and blood samples
See Appendix MethodsS1.

| Data analysis for determining DEGs
The differential gene expression analysis between different sample types was performed using the negative binomial statistical model of read counts as implemented in the edgeR Bioconductor package. 4 The cluster analysis of all genes differentially expressed in individual comparisons is performed using the Bayesian infinite mixture models. 5 The gene set enrichment analysis (GSEA) was performed using the LRpath methodology 6 as implemented in the CLEAN package. 7 The DEGs derived from sequenced data were compared between HAE (either Type 1 or HAEnCI) versus non-HAE controls and within the HAE patients between samples collected during flares and quiescence. These analytical methods attempted to identify biomarkers that could predict flares and disease progression, and thereby identify potential therapeutic agents to control such events.
The DEGs were queried within the Ingenuity knowledge base included in the Ingenuity Pathway Analysis® (IPA, Qiagen) software to perform the core analysis that permits interpretation of the biological relevance of expression changes in the omics data and to determine which metabolic and signaling canonical pathways were enriched in the active flare lesions in HAE Type 1 versus normal healthy skin from non-HAE controls. The core and subsequent comparison analyses using this application are summarized below.
Additionally, the regulator effects analysis, based on the comparison analysis, were also performed that explains the upstream regulators expected to affect the target molecules in the dataset.

| Association of PLAUR with severity of HAE angioedema attacks
Differential gene-expression of PLAUR in skin biopsies from HAE subjects at baseline versus non-HAE controls were regressed on the severity of angioedema attacks in HAE subjects using generalized linear regression assuming a log-normal distribution of gene expression values. Severity of attacks were determined clinically by the treating HAE specialist.

| Core analysis
Gene expression between comparison groups was explored by IPA® using the expression analysis component of the core analysis to determine the canonical pathways that were enriched in the data.
This analysis identified relationships, mechanisms, functions, and pathways relevant to the dataset. Subsequently, the significant upstream regulators affecting the molecules in the dataset were predicted. The biological functions significantly over-represented were determined. Molecular networks biologically relevant for HAE were determined based on the predicted activation or inhibition of regulatory molecules or biological processes. The p-value of overlap calculated using the right-tailed Fisher's Exact test with Benjamini-Hochberg correction for multiple testing, and the z-score, indicative of similar or dissimilar biological signatures between analysis for downstream effects or upstream regulators, were used for determining statistical significance.

| Comparison analysis
Comparison analysis was performed on the core analysis results from all the comparator groups to identify trends, similarities and differences between those results. Components of such comparison analyses include comparison of the canonical pathways, and the regulator effects analysis. Thus, the canonical pathways commonly represented in the core analysis between the comparison groups (i.e., the canonical pathways in the core analysis from HAE Type 1 flare vs. control, HAE Type 1 flare vs. HAEnCI, and others) were compared. Thereafter, upstream analysis compared the potential upstream regulators for the DEGs between the different comparison groups. Enriched biological and pathological functions based on these regulators and their target molecules were compared. Regulator effects analysis examined connections between upstream regulators, the dataset molecules and downstream functions in HAE pathogenesis, and determined if any specific therapeutic molecules could hypothetically minimize or prevent the differential gene expression thus preventing the pathogenesis of HAE flares.

| miRNA analysis from blood samples
The list of differentially expressed miRNA were uploaded to the IPA.
Core analyses for each comparison group were performed. Highly restrained stringent filters were used for querying only those molecules and relationships very relevant to humans. Pathway tools were used to generate and test additional hypothesis. The "build" tool was SINGH AND BERNSTEIN -3 of 11 used to determine which canonical or well-established pathways were associated with the molecules in the study data. Experimentally validated mRNAs, regulated by the differentially expressed miRNA in the data were determined. The canonical pathways, associated with the differentially expressed miRNA experimentally validated to regulate mRNA expression in TargetScan, miRecords, Ingenuity Expert Findings, were investigated. IPA tools were used to customize the networks and pathways that permitted generation of additional hypotheses from the networks identified by these tools.

| HAE Type 1 baseline and flare skin biopsy samples versus control samples
The PLAUR gene that encodes u-PAR was overexpressed at baseline and its expression further increased during flares in HAE-Type-1 patients compared to non-HAE and HAEnCI controls suggestive of overactive u-PAR-mediated signaling via binding to Factor-XII. 8 Increased baseline PLAUR expression in HAE Type I patients compared with non-HAE controls was associated with severe angioedema attacks (p = 0.0003), (Appendix Table 1). Using GSEA of the DEGs between HAE Type I baseline and flare samples, 18 DEGs were determined to be enriched in beta1/beta3integrin cell surface interactions and IL-6-mediated signaling events.

| Canonical pathways in core analysis
Appendix Tables 2 and 3 summarize the predicted canonical pathways using core analysis which were necessary for performing the comparison analysis.  Table 1 and Figure 3 show the upstream regulators that are pre- However, there is an established relationship between IL1β and several dataset molecules such as VEGFA, CXCL8 and TGFβ1. [9][10][11][12] In this study, IL1is shown to be significantly overexpressed in patients with HAE Type 1 versus non-HAE controls. As mentioned in Table 1,   Another potential limitation of this study could be the size of the groups analyzed. However, our initial power assumptions predicted F I G U R E 5 Network analysis of differentially expressed miRNA between HAE baseline versus HAE Type I flare blood samples. Canonical pathways associated with mRNAs regulated by differentially expressed miRNA (i.e., miR-184 and miR-1-3p) are shown. "Build" tools in IPA were used to determine the molecules associated with AKT2 (mRNA regulated by miR-184). HAE, hereditary angioedema statistically significant differences would be found between HAE

ACKNOWLEDGMENT
The investigator-initiated study was funded by Takeda (Shire) Pharmaceuticals.