Meta-analysis of Global and High Throughput Public Gene Array Data for Robust Vascular Gene Expression Discovery in Chronic Rhinosinusitis: Implications in Controlled Release

Abstract

Background:

Chronic inflammation is known to cause alterations in vascular homeostasis that directly affects blood vessel morphogenesis, angiogenesis, and tissue permeability. These phenomena have been investigated and exploited for targeted drug delivery applications in the context of cancers and other disease processes. Vascular pathophysiology and its associated genes and signaling pathways, however, have not been systematically investigated in patients with chronic rhinosinusitis (CRS). Understanding the interplay between key vascular signaling pathways and top biomarkers associated with CRS may facilitate the development of new targeted delivery strategies and treatment paradigms. Herein, we report findings from a gene meta-analysis to identify key vascular pathways and top genes involved in CRS.

Methods:

Proprietary software (Illumina BaseSpace Correlation Engine) and open-access data sets were used to perform a gene meta-analysis to systematically determine significant differences between key vascular biomarkers and vascular signaling pathways expressed in sinonasal tissue biopsies of controls and patients with CRS.

Results:

Thirteen studies were initially identified, and then reduced to five after applying exclusion principle algorithms. Genes associated with vasculature development and blood vessel morphogenesis signaling pathways were identified to be overexpressed among the top 15 signaling pathways. Out of many significantly upregulated genes, the levels of pro angiogenic genes such as early growth response (EGR3), platelet endothelial cell adhesion molecule (PECAM1) and L-selectin (SELL) were particularly significant in patients with CRS compared to controls.

Discussion:

Key vascular biomarkers and signaling pathways were significantly overexpressed in patients with CRS compared to controls, suggesting a contribution of vascular dysfunction in CRS pathophysiology. Vascular dysregulation and permeability may afford opportunities to develop drug delivery systems to improve efficacy and reduce toxicity of CRS treatment.

Introduction

Chronic Rhinosinusitis (CRS) is one of the most prevalent and debilitating chronic inflammatory diseases, affecting approximately 12% of the U.S. population [1]. In the last 15 years, therapeutic advancements have focused on the local delivery of topical corticosteroids through the development of drug-eluting stents and nasal aerosol delivery devices [2, 3]. Despite the improvements in topical delivery systems, approximately 20% of patients with CRS require surgical intervention to control their disease [4]. The development of new, safe, and targeted systemic delivery methods would greatly benefit those patients who fail topical medical management.

Chronic inflammation is known to cause alterations in vascular homeostasis that directly affects blood vessel morphogenesis, angiogenesis, and tissue permeability. Enhanced vascular permeability and hyperpermeability has been exploited in inflammatory pathologies for improved drug delivery to target tissue [5]. Further defining and characterizing vascular pathophysiology and its associated genes and signaling pathways, however, has not been systematically investigated in patients with CRS. Understanding the interplay between key vascular signaling pathways and top biomarkers associated with CRS may facilitate the development of new systemically delivered treatment paradigms, whereby the drug concentration is increased at the target site with reducing off-target accumulation.

Prior investigations have utilized gene array analysis as a tool to evaluate inflammatory gene expression patterns, identify potential new biomarkers, and define endotypes of subpopulations in individual cohorts of patients with CRS [6]. These studies have begun to shed light on many meaningful transcription products involved in CRS, including inflammatory cytokines and genes implicated in neoplastic processes, apoptosis inhibition, and fibrosis. Microarray analysis additionally enables genome-wide comparisons between disease and non-disease states. Recent advancements in bioinformatics have made microarray meta-analysis widely available through the use of proprietary software, enabling the comparison of key signaling pathways and biomarkers across gene array data sets to facilitate new qualitative and quantitative studies including pathophysiology-focused research investigations, biomarker discovery, and improved drug development. [7].

In this manuscript we report a meta-analysis on publicly available gene-array biosets to characterize the interplay between key vascular signaling pathways and top biomarkers associated with CRS. To our knowledge, this is the first meta-analysis that specifically examines large sample analysis of vascular gene expression in CRS. We hypothesized that patients with CRS will demonstrate unique vascular pathways and associated genes compared to patients without CRS.

Materials and Methods:

Software and apps

Illumina BaseSpace Correlation Engine (San Diego, CA) was used to run this meta-analysis study [8]. The correlation engine software has several applications that allow the user to derive a consensus gene signature and/or discover sets of commonly regulated biogroups based on a collection of individual biosets (Figure 1). Correlation Engine filters the top results based on overall statistical significance and consistency of the enrichment, or overlap, between the set of genes. Combination of these genes makes up each biogroup within a selected curated study for the meta-analysis. Based on these factors, the most significant biogroup or gene is given a score of 100, and all other biogroups and genes are normalized to the top-ranked biogroup or gene. The top 15 biogroups were first selected based on their overall statistical significance. The top upregulated and downregulated genes were then selected, and their biological roles and associated signaling pathways in the human body were identified.

Figure 1: Screenshot of BaseSpace software used for meta-analysis.

The solid red box represents the software. The dashed red box represents the different apps that are available for use, such as meta-analysis and body atlas. The dotted red box highlights the 5 biosets that were identified after exclusion criteria was applied.

Inclusion and exclusion criteria

The inclusion-exclusion principle is a combinatorial method to include and determine the size of a bioset or the probability of inclusion and exclusion based on a specific query. An inclusion criteria algorithm was employed to search for previously performed and publicly available studies that investigated the genomic variations in CRS via a microarray or RNA-seq platform. The terms ‘Chronic Rhinosinusitis’ and ‘Chronic sinusitis’ were searched as the primary boundary condition. These terms were selected on the basis of their definition: ‘chronic inflammation of the mucous membrane in the nose.’ A total of 13 studies were identified after searching for ‘Chronic Rhinosinusitis’ and ‘Chronic sinusitis’. An exclusion principle algorithm was then employed to narrow down the studies to fit our selection criteria, which is explained further in Figure 2. The secondary boundary condition for the exclusion principle was set to compare data generated from only sinonasal biopsies collected from patients with CRS vs. healthy or non-disease control tissues, and not any other type of tissue, mucus or blood.

Figure 2: Exclusion criteria.

Flow chart depicting the number of biosets that were identified and the criteria that were applied to reject studies that did not fit the scope of our meta-analysis objective.

The total number of 13 studies were identified after the application of the primary and secondary boundary conditions in the Illumina Correlation Engine. We then filtered the studies based on type of organism: homo sapiens, to avoid any any discrepancies arising from the organism type. This filtered the total number of studies down to 12 as one of the gene array studies was perfomed in sinonasal tissue obtained from mouse. Further, we selected only those studies that offered RNA expression results and discarded any studies comparing somatic mutations or DNA copy numbers to avoid correlations or comparisons of different genetic material, which can result in false data interpretation. To further select a more homogenous CRS population, phenotypic variations and/or variation in cell type in CRS were excluded from the analysis. Patients with other heterogenous causes of CRS such as aspirin-exacerbated respiratory disease, cystic fibrosis or studies comparing nucleophilic CRS vs eosinophilic CRS were excluded in this meta-analysis, to avoid cross-contamination between different phenotypes of CRS. This rendered the total number of studies available for comparison to 5. The Correlation Engine named each selected curated study for meta-analysis as a ‘Bioset’. . After applying the exclusion principle algorithm, 5 biosets were identified for meta-analysis (Figure 2 and Table 1), which were further manually evaluated for inclusion and exclusion criteria fit.

Table 1: Five identified biosets for meta-analysis.

The numbers of genes reported in each bioset are also listed.

Bioset #	Description	Upregulated genes	Downregulated genes	Reference
1	CRS vs healthy patients	2936	3825	[47]
2	CRS vs healthy patients	590	421	[48]
3	CRS vs healthy patients	3756	3874	[49]
4	CRS vs healthy patients	856	1028	[49]
5	CRS vs healthy patients	1008	1576	[49]

Bioset summaries

The bioset summaries were copied from Illumina BaseSpace Correlation Engine by clicking on the ‘Bioset summary’ link for each study.

Statistics:

BaseSpace Correlation Engine uses a proprietary algorithm: Fisher running algorithm and the Fisher exact test [8–12]. These algorithms are used to compute statistical significance between different biosets and genes that are differentially expressed. The most important two parameters in each bioset are the activity level of a gene (fold change) and the number of biosets in which that particular gene is active. The most significant gene (combination of fold change and p-values) was given a score of 100, and subsequent significant genes were normalized and given a score out of 100.

Results:

Selected biosets

After screening the identified biosets using inclusion and exclusion principle algorithms, the top 5 biosets shown below (1–5) were identified and manually assessed to ensure inclusion and exclusion criteria fit. The number of genes that were studied or reported by each respective research group was also identified (Table 1).

1. CRS Patients vs_ non-CRS patients

Study: Patients with CRS, Series GSE10406

Analysis summary: Genes with statistically significant differences between test and control conditions. Statistical analyses were performed on log-scale data.

Statistical tests: Parametric tests were performed, assuming unequal variances (Welch t-test) and a p-value cutoff less than 0.05. An additional fold change of 1.2 was applied to generate the final list of genes.

Gene filtering: Genes that had a mean normalized test and control intensity falling below the 20^th percentile of the combined normalized signal intensities were removed. The intensities reported below correspond to the mean normalized test and control data rescaled to a median of 500.

Platform: Affymetrix GeneChip Human HG_U133 Plus 2.0

2. CRS patients _vs_ non-CRS patients

Study: Patients with CRS, Series GSE72713

Analysis summary: Reads are aligned to the human genome (UCSC iGenomes hg19, download date 5/23/2014) using STAR 2.3 and RefSeq annotations. Reads are assigned to a gene if the read (or both reads in a pair) uniquely and fully map to the exons of one gene. The differential gene expression between the control and test sample groups were generated based on these counts using DESeq2. The base-mean read count, fold change, p-value, and q-value (Benjamini-Hochberg adjusted) were derived from this analysis. The median FPKMs per group were calculated separately based on normalized read counts, number of aligned reads, and the full gene length. The genes were filtered with a q-value cutoff of less than 0.05. An additional fold change cutoff of +/−1.2 was applied to generate the final list of genes.

Platform: Illumina iGenome UCSC, hg19, March 6, 2013 RefSeq

3. CRS patients _vs_ non-CRS patients

Study: Patients with CRS, Series GSE36830

Analysis summary: Genes with statistically significant differences between control and test samples. Statistical analyses were performed on log scale data.

Statistical tests: Parametric tests were performed, assuming unequal variances (Welch t-test) and a p-value cutoff less than 0.05. An additional fold change of 1.2 was applied to generate the final list of genes.

Gene filtering: Genes that had a mean normalized test and control intensity falling below the 20^th percentile of the combined normalized signal intensities were removed. The intensities reported below correspond to the mean normalized test and control data rescaled to a median of 500.

Platform: Affymetrix GeneChip Human HG_U133 Plus 2.0

4. CRS patients vs non-CRS patients

Study: Patients with CRS, Series GSE36830

Analysis summary: Genes with statistically significant differences between test and control conditions. Statistical analyses were performed on log-scale data.

Statistical tests: Parametric tests were performed, assuming unequal variances (Welch t-test) and a p-value cutoff less than 0.05. An additional fold change of 1.2 was applied to generate the final list of genes.

Gene filtering: Genes that had a mean normalized test and control intensity falling below the 20^th percentile of the combined normalized signal intensities were removed. The intensities reported below correspond to the mean normalized test and control data rescaled to a median of 500.

Platform: Affymetrix GeneChip Human HG_U133 Plus 2.0

5. CRS patients vs non-CRS patients

Study: Patients with CRS, Series GSE36830

Analysis summary: Genes with statistically significant differences between test and control conditions. Statistical analyses were performed on log-scale data.

Statistical tests: Parametric tests were performed, assuming unequal variances (Welch t-test) and a p-value cutoff less than 0.05. An additional fold change of 1.2 was applied to generate the final list of genes.

Gene filtering: Genes that had a mean normalized test and control intensity falling below the 20^th percentile of the combined normalized signal intensities were removed. The intensities reported below correspond to the mean normalized test and control data rescaled to a median of 500.

Platform: Affymetrix GeneChip Human HG_U133 Plus 2.0

Top 15 selected signaling pathways and biological roles

As a result of bioset analysis, multiple signaling pathways were determined to be overexpressed (Figure 3 and Table 2), with the most relevant signaling pathways being related to immune response activation, including ‘regulation of cell activation’ and ‘innate immune response.’ Among the top 15 selected pathways, a large number of genes associated with ‘vascular development’ and ‘blood vessel morphogenesis’ were found to be significantly different between controls and patients with CRS, suggesting the importance of these pathways in CRS vascular pathophysiology.

Figure 3: List of the top 15 overexpressed signaling pathways in CRS, as analyzed by meta-analysis of 5 gene array biosets.

The top 15 signaling pathways are shown in decreasing order of significance starting from the top. The size of circle represents the number of genes studied in each respective bioset (For scale, smallest circle represents 11 genes and the largest circle represents 164 genes).

Table 2: Top 15 signaling pathways identified.

The top 15 signaling pathways and their biological roles. The score given to each pathway is based on Correlation’s Engine statistical analysis.

Biogroup	Score	Role in the human body
Regulation of cell activation	100	Modulates the frequency, rate or extent of cell activation, the change in the morphology or behavior of a cell resulting from exposure to an activating factor such as a cellular or soluble ligand [50]
Innate immune response	99	Defense responses mediated by germline encoded components that directly recognize components of potential pathogens [51]
Positive regulation of immune response	96	Activates or increases the frequency, rate or extent of the immune response, the immunological reaction of an organism to an immunogenic stimulus [52]
Regulation of lymphocyte activation	89	Any process that modulates the frequency, rate or extent of lymphocyte activation [53]
Inflammatory response	87	The immediate defensive reaction (by vertebrate tissue) to infection or injury caused by chemical or physical agents. The process is characterized by local vasodilation, extravasation of plasma into intercellular spaces and accumulation of white blood cells and macrophages [16]
Leukocyte activation	85	A change in morphology and behavior of a leukocyte resulting from exposure to a specific antigen, mitogen, cytokine, cellular ligand, or soluble factor [54]
Lymphocyte activation	75	Change in morphology and behavior of a lymphocyte resulting from exposure to a specific antigen, mitogen, cytokine, chemokine, cellular ligand, or soluble factor [55]
External side of plasma membrane	74	The side of the plasma membrane that is opposite to the side that faces the cytoplasm [56]
Immunoglobulin subtype	72	The basic structure of immunoglobulin (Ig) molecules is a tetramer of two light chains and two heavy chains linked by disulphide bonds [57]
Vasculature development	72	The process whose specific outcome is the progression of the vasculature over time, from its formation to the mature structure. [58]
Activation of immune response	69	Any process that initiates an immune response [59]
Leukocyte migration	68	The movement of a leukocyte within or between different tissues and organs of the body [60]
Cell surface	68	The external part of the cell wall and/or plasma membrane [61]
Regulation of defense response	68	Any process that modulates the frequency, rate or extent of a defense response [62]
Blood vessel morphogenesis	64	The process in which the anatomical structures of blood vessels are generated and organized [63]

‘Vasculature development’ and ‘Blood vessel morphogenesis’ pathways

We compared the significance of the ‘vasculature development’ and ‘blood vessel morphogenesis’ pathways amongst all of the biosets (Figure 4). Both signaling pathways were found to be significant across all 5 biosets (confidence threshold set to 95%). We then identified the top 10 genes that were significantly differentially expressed in each bioset with respect to ‘vasculature development’ (Table 3) and ‘blood vessel morphogenesis’ (Table 4) pathways. Several genes such as early growth response 3 (EGR3), transforming growth factor beta 2 (TGFB2), 15-hydroxyprostaglandin dehydrogenase (HPGD) and collagen and calcium binding EGF domains 1 (CCBE1) were found to be significantly elevated in more than one bioset.

Figure 4. Significance of signaling pathways.

(A) Significance of ‘vasculature development’ across different biosets. (B) Significance of ‘blood vessel morphogenesis’ across different biosets.

Table 3:

The top 10 differentially expressed genes in each bioset within the ‘vasculature development’ pathway.

Bioset	Top 10 differentially expressed genes associated with vasculature development
1	SFRP1, ERRFI1, HPGD, EGR3, PLAT, ENPEP, ZFAND5, RHOB, CEACAM1, NR4A1
2	SLIT2, CCBE1, IL8, SFRP4, ADM, SAT1, PDE3B, SERPINE1, SOCS3, TYMP
3	SCG2, EGF, NKX3–1, CCL2, THY1, NDP, HMOX1, CCBE1, THBS1, GJC1
4	CITED2, HPGD, EGR3, HPGD, APOLD1, NOX1, NDP, CXCL12, TGFB2, PECAM1
5	CCBE1, SCG2, HPGD, TGFB2, BMPER, SFRP2, FGF9, ITGB1, MYLK, PECAM1

Table 4:

The top 10 differentially expressed genes in each bioset within the ‘blood vessel morphogenesis’ pathway.

Bioset	Top 10 differentially expressed genes associated with blood vessel morphogenesis
1	HPGD, EGR3, PLAT, ENPEP, RHOB, CEACAM1, NR4A1, TIPARP, TFAP2B, VEGFA
2	SLIT2, CCBE1, IL8, ADM, SAT1, PDE3B, SERPINE1, TYMP, WARS, FGF9
3	SCG2, EGF, CCL2, THY1, HMOX1, CCBE1, THBS1, GJC1, GJA5, GREM1
4	CITED2, HPGD, EGR3, APOLD1, NOX1, CXCL12, TGFB2, HHEX, NOTCH3, MCAM
5	CCBE1, SCG2, HPGD, TGFB2, BMPER, TGFB2, SFRP2, FGF9, ITGB1, MYLK

Top upregulated and downregulated genes

We analyzed the top up - and downregulated genes identified in our meta-analysis and screened for associations to vasculature-related signaling pathways (Tables 5 and 6). The top overexpressed genes implicated in pro-angiogenic and vasculature signaling pathways included: parathyroid hormone-like hormone (PTHLH), CD27 molecule (CD27), carcinoembryonic antigen-related cell adhesion molecule (CEACAM21), platelet/endothelial cell adhesion molecule (PECAM1), ADAM metallopeptidase domain (ADAM19), and L-Selectin (SELL). The top downregulated genes related to pro-angiogenic and vasculature signaling pathways included: solute carrier family 6 (SLC6), coiled coil domain (CCD), epidermal growth factor-like (EGFL), and mucin 15 (MUC15).

Table 5: The top upregulated genes.

The gene abbreviation is followed by a score which is provided by the BaseSpace correlation engine. The descriptive name of the gene is also mentioned. Corresponding role of the gene in our body is also listed.

	Gene	Score	Description	Role in the human body	Implicated in
					Inflammatory diseases	Drug delivery
1	CCL19	94	Chemokine (C-C motif) ligand 9	Chemokines are family of secreted proteins involved in immunoregulatory and inflammatory processes [64]	Yes [64]	Yes [65]
2	NOX2	83	NADPH oxidase 2	Primary component of the microbicidal oxidase system of phagocytosis [66]	Yes [67]	Yes [67]
3	FBLN1	82	Fibulin 1	It is a secreted glycoprotein that mediates platelet adhesion via binding fibrinogen [68]	Yes [69]	Yes [70]
4	HLA-DPA1	80	Major histocompatibility complex, class II, DP alpha 1	Plays a central role in the immune system by presenting peptides derived from extracellular proteins [71]	NA	NA
5	TMEM176B	76	Transmembrane protein 176B	Protein coding gene that plays a role in the maturation of dendritic cells [72]	NA	Yes [73]
6	LST1	75	Leukocyte specific transcript 1	Can inhibit the proliferation of lymphocytes [74]	Yes [75]	NA
7	PTHLH	75	Parathyroid hormone-like hormone	Regulates endochondral bone development and epithelial-mesenchymal interactions during the formation of mammary glands and teeth [76]	Yes [77]	NA
8	TDO2	71	Tryptophan 2,3-dioxygenase	Plays a role in catalyzing the first and rate-limiting step in major tryptophan metabolism [78]	Yes [79]	Yes [79]
9	CCL13	69	Chemokine (C-C motif) ligand 13	Cytokines are family of secreted proteins involved in immunoregulatory and inflammatory processes. Responsible for accumulation of leukocytes during inflammation [80]	Yes[81]	NA
10	CD79A	66	CD791 molecule, immunoglobulin-associated alpha	It is a B lymphocyte antigen receptor. It includes the antigen-specific component, surface immunoglobulin (Ig). It encodes the Ig-alpha protein of the B-cell antigen component [82]	NA	NA
11	CSF2RB	66	Colony stimulating factor 2 receptor	Defects in this gene have been reported to be associated with protein alveolar proteinosis (PAP) [83]	NA	NA
12	IGHG3	66	Immunoglobulin heavy constant gamma 3	Membrane-bound immunoglobulins serve as receptors which, upon binding of a specific antigen, trigger the differentiation of B lymphocytes into immunoglobulins-secreting plasma cells [84]	Yes [85]	NA
13	CLEC7A	65	C-type lectin domain family 7	It functions as a pattern-recognition receptor that recognizes a variety of beta-1,3-linked glucans from fungi and plants, and in a way plays a role in innate immune response [86]	NA	NA
14	CD27	63	CD27 molecule	Plays a role in regulating B-cell activation and immunoglobulin synthesis [87]	Yes [88]	Yes [89]
15	RNASE2	62	Ribonuclease, RNAse family 2	Protein has antimicrobial activity against viruses [90]	NA	NA
16	EMR1	62	Egf-like module containing, mucinlike, hormone receptor like 1	EMR1 expression in humans is restricted to eosinophils and is a specific marker for these cells [91]	Yes[41]	Yes[41]
17	FCER1G	62	Fc fragment of IgE	The high affinity IgE receptors is a key molecule involved in allergic reactions [92]	Yes [93]	NA
18	ZBP1	61	Z-DNA binding protein 1	It encodes a Z-DNA binding protein. Z-DNA formation is a dynamic process, largely controlled by the amount of supercoiling [94]	Yes [95]	NA
19	CEACAM21	60	Carcinoembryonic antigen-related cell adhesion molecule 21	Encoded proteins mediate cell adhesion and play a role in arrangement of tissue threedimensional structure, angiogenesis, apoptosis, metastasis and modulation of innate immune response [96]	NA	NA
20	ADAM19	59	ADAM metallopeptidase domain 19	It serves a s a marker for dendritic cell differentiation. They are also involved in normal physiological and pathological processes such as cell migration, cell adhesion, signal transduction etc. [97]	Yes [97]	Yes [97]
21	H2BC8	59	Histone cluster 1, H2bg	Plays a role in Cellular senescence, chromosome maintenance, cell cycle etc. [98]	NA	NA
22	SLAMF7	58	SLAM family member 7	It is a robust marker of normal plasma cells and malignant plasma cells in multiple myeloma [99]	NA	Yes [100]
23	SLFN5	57	schlafen family member 5	It is a member of the Schlafen family and may have a role in hematopoietic cell differentiation [101]	NA	NA
24	AIF1	57	Allograft inflammatory factor 1	Involved in negative regulation of growth of vascular smooth muscle cells, which contributes to the antiinflammatory response to vessel wall trauma [102]	NA	NA
25	PECAM1	55	Platelet/endothelial cell adhesion molecule 1	Makes up a large portion of endothelial cell intercellular junctions and is involved in angiogenesis, vasculature development, leukocyte migration etc. [26]	Yes [28]	Yes [103]
26	HLA-DMB	54	Major histology complex, class II	Plays a central role in the peptide loading of MHC class II molecules by helping to release class II-associated invariant chain peptide molecule from the peptide binding site [104]	Yes [46]	Yes [46]
27	CYSLTR1	53	Cysteinyl leukotriene receptor 1	Activation of this receptor by LTD4 results in contradiction and proliferation of smooth muscle, edema, eosinophil migration and damage to the mucus layer in the lung [105]	Yes [106]	Yes [106]
28	SELL	53	Selectin L	Plays an important role in leukocyte-endothelial cell interactions [27]	Yes [107]	Yes [107]
29	ARHGDIB	51	Rho GDP dissociation inhibitor beta	Involved in diverse cellular events, including cell signaling, proliferation, cytoskeletal organization and secretion [108]	NA	NA
30	NCF4	51	Neutrophil cytosolic factor 4	Pi3 kinase signaling pathways are activated that are a part of cellular functions such as cell growth, proliferation, differentiation, survival etc. [109]	Yes [110]	NA

Table 6. The top downregulated genes.

The gene abbreviation is followed by a score which is provided by the BaseSpace correlation engine. The descriptive name of the gene is also mentioned. Corresponding role of the gene in our body is also listed.

	Gene	Score	Description	Role in the human body	Implicated in
					Inflammatory disease	Drug Delivery
1	SEC14L3	100	SEC-14 like 3	Essential for biogenesis of Golgi-derived transport vesicles, and thus is required for the export of yeast secretory proteins form the Golgi complex [111]	Yes [112]	NA
2	SLC6A13	92	Solute carrier family 6, member 13	Involved in signaling pathways such as, synaptic vesicle cycle and benzodiazepine pathway [113]	NA	NA
3	FYB2	92	Chromosome 1 open reading frame 168	Plays a role in T-cell receptor mediated activation of signaling pathways. Required for T-cell activation and integrin-mediated T-cell adhesion in response to TCR simulation [114]	NA	NA
4	GSTA3	89	Glutathione S-transferase alpha 3	Involved in cellular defense against toxic, carcinogenic and pharmacologically active electrophilic compounds [115]	Yes [116]	NA
5	OSBPL6	85	Oxysterol binding protein-like 6	The gene encodes a member of the oxysterol-binding protein (OSBP) family, a group of intracellular lipid receptors and the related pathways are synthesis of bile acids and bile salts and metabolism [117]	NA	NA
6	CCDC81	84	Coiled-coil domain containing 81	This gene is a member of the coiled coil domain structure family. Other functions are still unknown [118]	NA	NA
7	HHLA2	83	HERV-H LTR associating 2	Thought to regulate cell-mediated immunity by binding to a receptor on T lymphocytes and inhibiting proliferation of these cells [119]	NA	Yes [120]
8	ZBTB16	79	Zinc finger and BTB domain containing 16	This gene is involved in cell cycle progression, and interacts with a histone deacetylase [121]	NA	NA
9	SLC15A2	79	Solute family carrier 15, member 2	Responsible for the absorption of small peptides, as well as beta-lactam antibodies and other peptidelike drugs, from the tubular filtrate [122]	Yes [123]	NA
10	DNAH5	79	Dynein, axonemal, heavy chain 5	Functions as a force-generating protein with ATPase activity, whereby the release of ADP is thought to produce the force-producing power stroke [124]	NA	NA
11	CSMD1	76	CUB and Sushi multiple domains 1	It is believed that the gene product of CSMD1 functions as Complement control protein [125]	NA	NA
12	CES1	76	Carboxylesterase 1	Responsible for hydrolysis of ester- and amide-bond containing drugs such as cocaine and heroin [126]	NA	NA
13	ARX	76	Aristaless related homebook	This gene is thought to be involved in Central Nervous System (CNS) development. Mutations in these genes cause epilepsy [127]	NA	NA
14	EGFL6	74	EGF-like-domain, multiple 6	Members of this superfamily (Epidermal Growth Factor family) are often involved in the regulation of cell cycle, proliferation and development processes [128]	NA	Yes [129]
15	KLHL32	73	Kelch-like 32	No reported function for this gene so far	NA	NA
16	TTC6	72	Tetratricopeptide repeat domain 6	Has shared roles in cilia formation and function [130]	NA	NA
17	ERICH5	72	Chromosome 8 open reading frame 47	Encodes a ubiquitously expressed protein of unknown function. It co-localizes at the base of primary cilium in human retinal pigment epithelial cells [131]	NA	NA
18	ARG2	71	Arginase, type II	They are NADPH-dependent flavoenzymes that catalyzes the oxidation of soft nucleophilic heteroatom centers in drugs, pesticides and xenobiotics [132]	Yes [133]	NA
19	FMO5	70	Flavin containing monooxygenase 5	Flavin-containing monooxygenases are NAPDH-dependent flavoenzymes that catalyzes the oxidation of soft nucleophilic heteroatom centers in drugs, pesticides and xenobiotics [134]	Yes [135]	NA
20	SLC23A1	69	Solute carrier family 23, member 1	The encoded protein is active in bulk vitamin C transport involving epithelial surfaces [136]	Yes [137]	NA
21	VEPH1	69	Ventricular zone expressed PH domain homolog 1	Interacts with TGF-beta receptor type-1 (TGFBR1) and inhibits dissociation of activated SMAD2, impeding its nuclear accumulation and resulting in impaired TGF-beta signaling [138]	NA	NA
22	NELL2	67	NEL-like 2 (chicken)	Plays a role in cell growth and differentiation as well as in oncogenesis [139]	NA	NA
23	MUC15	67	Mucin 15, cell surface associated	May play a role in the cell adhesion to the extracellular matrix [140]	Yes [141]	Yes [142]
24	PKIB	67	Protein kinase inhibitor beta	Study suggests that this protein may interact with the catalytic subunit of cAMP-dependent protein kinase and act as a competitive inhibitor [143]	NA	NA
25	FOXP2	66	Forkhead box P2	The product of this gene is thought to be required for proper development of speech and language regions of the brain during embryogenesis [144]	NA	Yes [144]
26	DLG2	66	Discs, large homolog	The encoded protein may interact at postsynaptic sites to form a multimeric scaffold for the clustering of receptors, ion channels and associated signaling proteins [145]	NA	NA
27	ADRB1	64	Adrenoceptor beta 1	Mediates the physiological effects of hormone epinephrine and the neurotransmitter norepinephrine [146]	NA	NA
28	PER1	63	Period homolog 1	Plays a role in the modulation of the neuroinflammatory state via the regulation of inflammatory mediators release, such as CCL2 and IL6 [147]	Yes [148]	NA
29	ALDH3B1	63	Aldehyde dehydrogenase 3 family	Plays a major role in the detoxification of the aldehydes generated by alcohol metabolism and lipid peroxidation [149]	NA	NA
30	ERBB4	63	V-erb-a erythroblastic leukemia viral oncogene homolog 4	The protein binds to and is activated by neuregulins and other factors and induces a variety of cellular responses [150]	Yes [150]	Yes [151]

Discussion

The vascular system is critical for supplementing tissue with nutrients and proteins such as growth factors, as well as for removing waste products [13, 14]. An increase in vascular permeability is typically observed in inflammatory conditions [15] with associated increase in inflammatory gene expression [16]. This permeability is regulated by various vascular-permeabilizing agents, as well as vessel morphogenic and growth factors such as vascular endothelial growth factor (VEGF) and vascular endothelial growth factor A (VEGFA) [17]. Understanding the underlying vascular gene and pathway expression in CRS will allow for the strategic design and development of new nanoscale drug delivery technologies that exploit endothelial junction permeability to increase the extravasation of drug into the target sinonasal tissues.

Large sample studies are critical for studying the effects of genes on the development and regulation of the underlying vascular changes that occur in disease and performing meta-analysis on prior gene arrays allows this insight into CRS. This is in contrast to the current available level of evidence in gene expression studies which are based off of cohorts from single institutions with small sample sizes [18, 19]. Gene-array meta-analysis offsets this limitation by examining a larger population of patients at multiple institutions. Meta-analysis also helps to address the heterogeneity and biases that arise based on regional factors and differences such as patient demographics and environment [20]. Performing an analysis of the top genes that are over- and under-expressed affords a more comprehensive view of which genes are contributing to the neoangeogenesis that is occuring in patients with CRS [21, 22]. Further, key pathways and genes identified hold the promise of becoming potential therapeutic targets or biologic markers to monitor treatment responses for drug delivery platform development. Altogether, meta-analysis provides a more precise estimate of gene expression integration by reducing heterogeneity of the overall estimate, and ultimately, a reduction in the effects of individual study-specific biases.

Herein, we observed two vascular pathways (vasculature development and blood vessel morphogenesis) to be highly overexpressed in patients with CRS. Typically in adults these pathways are quiescent, with little to no vascular growth or remodeling, unless stimulated by inflammation or injury. These two pathways incorporate key genes that mediate neoangiogenesis, by stimulating endothelial cells to dynamically assemble into new blood vessels via growth factor-mediated signaling through VEGF that is known to strongly upregulate EGR3 [23]. EGR3 is also critical in the regulation of inflammation and antigen-induced proliferation of both B and T cells [24] that are associated with the pro-inflammatory response seen in patients with CRS. EGR3 is also hypothesized to be upregulated in both mice and humans to control inflammation [25]. Herein, we demonstrate EGR3 gene expression to be significantly upregulated in multiple biosets associated with the vasculature development and blood vessel morphogenesis in addition to its upregulation across all 5 biosets that were explored for this meta-analysis (Table 3, 4 and 5).

Leukocytye trafficking and endothelial tight junctions are two functions that are critical for controlling the inflammatory response in patients with CRS by regulating the permeability between endothelial cells to selectively allow the migration of leukocytes. Unregulated, endothelial cell remodeling results in an increase in the transmigration of inflammatory leukocytes, which in turn, release stimulatory inflammatory cytokines initiating further neoangiogenesis, perpetuating the inflammatory neo-angiogenesis cycle. Herein, we demonstrate significant overexpression of PECAM1 and SELL (Table 3–5) two key mediators involved in endothelial cell permeability (Table 5). PECAM1 and SELL function as regulators of leukocyte trafficking, and in the maintenance of endothelial cell junction permeability [26, 27]. As expected PECAM1 and SELL expression are preferentially upregulated and highly expressed in patients with CRS compared to controls [28]. Reducing leukocyte trafficking into the sinonasal mucosa and controlling permeability presents a novel unique target to further study.

It is clear that neoangiogenesis is a complex coordinated event that regulates blood flow, while regulating the activation of leukocytes and by selectively controlling the exchange of macromolecules [29]. However, due to the complex coordination of neoangiogenesis during chronic inflammation, this tightly controlled system results in leaky and dysfunctional blood vessels, allowing for an increased exchange of macromolecules. Interestingly, neoangiogenesis is also suggested to be central to the mechanism by which acute inflammation which initially can trigger the activation of endothelial cells can transition to chronic inflammation [30]. Although limited, our results align with prior investigations that have demonstrated blood vessel dysfunction, endothelial damage to the endothelial lining and widening of endothelial cell gap junctions, mediated through VEGF, EGR3, Tumor Necrosis Factor (TNF), and TGFβ in patients with CRS [31]. Further, vasculature development and blood vessel morphogenesis is well known to be involved in other inflammatory diseases such as rheumatoid arthritis (RA) that have inflammation-driven vascular changes [32].

Taking advantage of inflammation-driven vascular changes for nanoparticle drug delivery has not been explored for CRS but has been explored in inflammatory joint conditions such as RA [33, 34]. In fact, a large amount of effort and resources have been aimed at understanding and exploring ways to take therapeutic advantage of enhanced permeability and retention in disease [33, 35]. Several key landmark articles expand this concept to chronic inflammation in animal models of RA, for which the mechanism of Extravasation via Leaky Vasculature and subsequent Inflammatory cell-mediated Sequestration (ELVIS) was first introduced [36–38]. The authors, demonstrated that N-(2-hydroxypropyl)methacrylamide (HPMA) copolymer-dexamethasone conjugates administered via intravenous injection in rats preferentially accumulated in the inflamed joints to facilitate the reduction of arthritis-associated inflammation. Similarly, Metselaar et al [39] demonstrated prolonged accumulation of glucocorticosteroid-encapsulated long-circulating liposomes, resulting in the complete remission of joint arthritis. In another study Gawne et al. looked at the correlation between the ankle and wrist joint swelling and uptake of PEGylated liposomal methylprednisolone liposomes labelled with radioisotope ⁸⁹Zr (NSSL-MPS). A significant correlation was identified between joint swelling (inflammation) and uptake of NSSL-MPS, which was not the case for free ⁸⁹Zr [40]. Similarly, defining and characterizing the vascular changes in CRS will allow investigators to design and deliver nano-drug delivery systems to promote increased efficacy while taking advantage of nano-drug delivery systems to improve targeting while reducing the off-target toxicities of free drug.

We next sought to better understand how our meta-analysis data could be applied to drug delivery. A comprehensive literature search was performed on all of genes found to be significantly up- or down-regulated in the meta-analysis. We sought to identify those genes that have already been implicated in both inflammation and drug delivery. This revealed 11 upregulated genes (Table 5) and 2 downregulated genes (Table 6) that have been linked to both inflammation and drug delivery. Specifically, genes we identified have been associated with targeting therapies (EMR1:[41], PECAM1:[42]), prognostic biomarkers for therapeutic efficacy (PECAM1:[43], SELL:[44]), defining vascular permeability and characterizing the pathophysiology of disease to improve drug design (PECAM1:[45]).

Biomarkers also provide understanding into disease progression and prognosis. Further, biomarkers are invaluable in drug delivery for designing efficacy studies in both pre-clinical and clinical trials. For example, Morel et al [46], demonstrate the advantage of HLA class II histocompatibility antigen, DM beta chain (HLA-DMB) as a novel prognostic biomarker in RA and correlated the expression of HLA-DMB to disease severity. Herein, we demonstrate significant upregulation of HLA-DMB across all 5 biosets of the meta-analysis study (Table 5). Such biomarkers have not been identified in patients with CRS. These findings may act as a starting point to find those differentially expressed genes in CRS that may be similarily used as biomarkers in efficacy studies. Additional studies are needed to further investigate and identify new biomarkers that can be used in the development of improved targeted drug delivery systems.

Although this investigation aims to diminish the limitations of single institution array investigations it is not without its own limitations. We acknowledge gene expression results of our meta-analysis could not be cross validated using other available platforms. Due to the limited number of acceptable array studies to include in the final meta-analysis, the data presented has inherent heterogeneity that should be considered. For example, each individual gene array looks at different types and different number of genes, making a direct comparison between two studies difficult.

Conclusions

A meta-analysis approach using publicly available gene array data enables the evaluation of differentially expressed genes and signaling pathways in disease. In our meta-analysis, we identified key signaling pathways and associated genes in patients with CRS and highlight the key genes associated with vasculature development, blood vessel morphogenesis, and angiogenesis with links to drug delivery. We report significant differential expression of genes associated with vascular signaling pathways such as EMR1, PECAM1 and SELL related to vascular pathophysiology in CRS. Additional studies should be focused on further defining vasculature development, blood vessel morphogenesis, and angiogenesis in CRS.

Highlights.

• Understanding the underlying vascular gene and pathway expression in CRS will allow for the strategic design and development of new nanoscale drug delivery

• We observed two vascular pathways (vasculature development and blood vessel morphogenesis) to be highly overexpressed in patients with CRS

• Vascular dysregulation and permeability contribute to the CRS pathophysiology

• This provides opportunity to develop nanoscale drug delivery systems to improve efficacy and reduce toxicity of CRS treatment