RNA Sequencing Analysis of Intracranial Aneurysm Walls Reveals Involvement of Lysosomes and Immunoglobulins in Rupture

Introduction

Intracranial aneurysms are common with a prevalence of 3%.¹ The availability of noninvasive imaging techniques has increased, which coincided with an increase in incidental detection of aneurysms.² Although only a minority of aneurysms ruptures, the consequences of aneurysmal subarachnoid hemorrhage are enormous, because of the young age at which it occurs, and the high case fatality and morbidity.³ Preventive treatment of unruptured aneurysms carries a risk of complications.⁴ New treatment strategies, therefore, inhibiting the formation, growth, and rupture of aneurysms are needed. To develop such treatment strategies, we need more insight into the pathogenesis of aneurysm development and rupture. Studies investigating differences in gene expression in aneurysms versus controls and in ruptured versus unruptured aneurysms may identify genes and pathways involved in development and rupture of aneurysms. Previous studies were small and, used microarray techniques,^5–12 and extracranial instead of intracranial arteries as control.^6–9,11,12 Recent advances in sequencing methodology, such as next-generation sequencing-based RNA profiling methods, provide more accurate, sensitive, comprehensive, and reliable gene expression data than microarrays.¹³ We compared differential expression of genes of a large sample of intracranial aneurysm tissue samples (n=44) to a unique control sample of intracranial arteries (n=16), and compared ruptured (n=22) to unruptured (n=21) aneurysms tissue samples using RNA sequencing, to gain insight in the pathogenesis of the development and the subsequent rupture of intracranial aneurysms.

Methods

Patient Selection and Tissue Collection

Aneurysm tissue samples were excised after complete obliteration of the aneurysm with a clip in patients aged ≥18 years undergoing neurosurgical clipping of a ruptured, or unruptured saccular intracranial aneurysm at the department of Neurology and Neurosurgery of the University Medical Center Utrecht, The Netherlands between 2010 and 2013. As controls, a tissue sample of an intracranial cortical artery was obtained from the resected brain tissue of patients aged ≥16 years who underwent surgery because of intractable epilepsy. Tissue samples were snap frozen in liquid nitrogen (<1 minute after excision) and stored at −80°C until further use. We collected characteristics for all patients and for those with aneurysms also aneurysm characteristics, including the largest diameter of the lumen of the aneurysm and the location of the aneurysm from the available imaging studies (mostly computed tomographic angiography). This study was approved by the Institutional Review Board of the University Medical Center Utrecht. For further details see the Methods section in the online-only Data Supplement.

Library Preparation and RNA Sequencing

An initial quality check of the extracted RNA of the samples was performed by capillary electrophoresis and RNA quantification using the LabChip GX (PerkinElmer, Waltham, MA). Please see the Methods section in the online-only Data Supplement for further details. The median RNA quality score of the samples was 6.7 (range, 4.2–9.3) with a score of 7.2 (4.2–9.0) in controls, 6.7 (4.8–8.9) in ruptured aneurysm samples, and 6.3 (5.1–9.3) in unruptured aneurysm samples (Table I in the online-only Data Supplement). Sequence libraries were generated from 44 aneurysms (22 ruptured, 21 unruptured, and 1 with unknown rupture status) and 16 control samples using the TruSeq mRNA sample preparation kit from Illumina (San Diego, CA) using the Sciclone NGS Liquid Handler (Perkin Elmer). After an extra purification step of the libraries with the automated agarose gel separation system Labchip XT (PerkinElmer), 9 picomoles of the obtained cDNA fragment libraries were sequenced on an Illumina HiSeq2500 using default parameters (single read 1×100 bp) in randomly arranged pools of 10 or 11 samples. On an average, 14 209 239 reads were generated per sample. The quality of the sequencing data were good (Figure I in the online-only Data Supplement). The average alignment of the reads to the human reference genome (uniquely mapped reads) was 86% (range, 73%–90%). For further details, see the Methods section in the online-only Data Supplement.

Gene Expression Quantification

We used R version 3.1.0¹⁴ and Bioconductor (version 2.14) packages edge R (version 3.6.2)¹⁵ and limma (version 3.20.2)¹⁶ for analysis of the gene expression data (for more details see the Methods section in the online-only Data Supplement). A generalized linear model was used to test the null hypothesis that there is no differential expression of genes between analysis groups. Age and sex of the patients were added to the model. In the model comparing aneurysms and controls, we added the rupture status, to control for genes with strong expression differences between ruptured and unruptured aneurysms. Common and tagwise dispersion estimates were calculated with the Cox–Reid profile-adjusted likelihood method to be able to correct for the technical and biological variation when fitting the multivariate negative binomial model. A negative binomial generalized log-linear model, using the tagwise dispersion estimates, was fitted to the read counts for each gene and a genewise statistical test was performed for the given coefficient (either aneurysms to controls or ruptured to unruptured aneurysms). Then, a likelihood ratio test for the given coefficient in the model was performed. To correct for multiple testing, we calculated Benjamini Hochberg false discovery rates (FDR) and considered genes with an FDR-adjusted P value <0.05 differentially expressed. Although the genes with low counts (genes with <1 read per million in n of the samples, where n is the size of the smallest group of replicates) are not likely to show significant results in the differential expression analysis, they can still be biologically relevant for the disease in case such genes are only expressed in 1 condition and have 0 counts in the other. We therefore performed a subanalysis of the raw count data and selected those genes that had >200 counts versus 0 counts in aneurysms versus controls and ruptured versus unruptured aneurysms.

Validation of RNA Sequencing Data With Digital Droplet Polymerase Chain Reaction

Eight genes were selected for validation by selecting the 2 most biologically relevant top genes with overexpression and the 2 most biologically relevant top genes with underexpression in the aneurysms versus control tissue analysis and in the ruptured versus unruptured aneurysms analysis. Their expression was studied by digital polymerase chain reaction in the remaining extracted RNA of the samples included in the differential expression experiment (for more details see the Methods section and Figure II in the online-only Data Supplement). In 2 aneurysm samples, the amount of RNA left after RNA sequencing was insufficient leaving 20 ruptured, 21 unruptured aneurysm samples, and 16 control tissue samples for further analysis in this validation step. The relative difference in gene expression between aneurysms and controls or ruptured and unruptured aneurysms was analyzed with the Mann–Whitney U test. Genes with a P value of <0.05 were considered statistically significant differentially expressed.

In addition, we reviewed the existing literature investigating differences in gene expression in aneurysms for the 8 selected genes for further validation.^5–12 Furthermore, we compared the list of differentially expressed genes in our study to the full lists of differentially expressed genes in the previous studies,^5–12 when these lists were available online or could be obtained from the authors. The results of this comparison are shown in the Results section and Tables II and III in the online-only Data Supplement.

Functional Network Analysis

To identify the biological functional pathways that were significantly over-represented by differentially expressed genes between aneurysm and control tissue samples as well as between ruptured and unruptured aneurysm samples, we analyzed which Kyoto Encyclopedia of Genes and Genomes functional pathways, and which Gene Ontology categories were enriched in our data set. We used the Bioconductor package goseq (version 1.18.0),¹⁷ which enabled correction for gene length bias in the analysis. We used all genes with differential expression with an FDR-adjusted P value <0.05 as input for the analysis. Kyoto Encyclopedia of Genes and Genome pathways with an FDR-adjusted P value <0.05 were considered significantly enriched. Highly correlated terms were removed from the enriched Gene Ontology lists using the Gotrimming software (for further details see the online-only Data Supplement).¹⁸

Results

We analyzed the transcripts of 44 aneurysm biopsies (22 ruptured, 21 unruptured, and 1 with unknown rupture status) of 38 aneurysm patients (6 patients had two aneurysms treated) and 16 control biopsies of 38 patients with aneurysm and 16 control patients. Baseline characteristics of the patients and controls are shown in Table 1.

Aneurysm Versus Control Tissue

Differentially Expressed Genes

The differential expression analysis yielded 51 genes with overexpression in aneurysms compared with controls (Figure 1; Table IV in the online-only Data Supplement). These included the top 5 genes collagen type X (COL10A1), cartilage intermediate layer protein 2 (CILP2), 1 RNA gene affiliated with the long noncoding RNA class (ENSG00000206195), secreted frizzled-related protein 2 (SFRP2), and muscle excess 3 RNA-binding family member B (MEX3B; Table 2). We found 178 genes that were underexpressed in aneurysm tissue (Table IV in the online-only Data Supplement), including the top 5 genes family with sequence similarity 134, member B (FAM134B), a gene of the solute carrier family (SLC13A3) which code for transporter proteins in the cell membrane, a gene involved in coagulation (SERPIND1), the growth regulation by estrogen in breast cancer 1 gene (GREB1) and a gap junction protein (GJB6; Table 2). The overexpression of COL10A1 and CILP2 in aneurysm tissue, seeming the 2 most biologically relevant top genes, was confirmed in the validation experiment (P<0.0001), as was the underexpression of GJB6 (P<0.0001) and SERPIND1 (P=0.0015; Table 2; Figure 2). The 4 relevant and highly significant genes in aneurysms versus controls were not found differentially expressed in the 4 previous studies of which a full lists of differentially expressed genes was available,^6–9 nor in the one in which an incomplete list was available.¹¹

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Functional Network Analysis

Functional network analysis of the 229 differentially expressed genes with an FDR-adjusted P value <0.05 did not identify any significant Kyoto Encyclopedia of Genes and Genome pathways. After removal of redundant classes, 51 Gene Ontology terms were identified (Table V in the online-only Data Supplement for the full list), including terms related to the extracellular matrix (ECM) and transmembrane transporter activity, and terms involving blood vessel regulation.

Low-Count Gene Analysis

The subanalysis of low-count genes in the raw count data yielded 3 immunoglobulin κ variable region genes (IGKV1D-42, IGKV3D-15, and IGKV1-6), 5 immunoglobulin heavy chain variable region genes (IGHV3-20, IGHV3OR16-15, IGHV3-60, IGHV1OR15-4, and IGHV3-66), and 2 other genes (tyrosinase [TYR] and a gene with Ensemble ID ENSG00000198229 but without an associated gene name) with >200 counts in aneurysm tissue and 0 counts in control tissue. None of the genes had 0 counts in aneurysms and >200 counts in control tissue.

Ruptured Versus Unruptured Aneurysm Tissue

Differentially Expressed Genes

The differential expression analysis identified 958 genes with overexpression in ruptured aneurysm tissue compared with unruptured aneurysm tissue (Figure 3; Table VI in the online-only Data Supplement). The top 5 overexpressed genes included the macrophage receptor with collagenous structure gene (MARCO), genes involved in ECM structure (transforming growth factor β-induced [TGFBI], heparanase [HPSE]), and 2 members of CD family (CD300C and CD300E; Table 2). Five hundred thirty-one genes were underexpressed in ruptured aneurysms (Table VI in the online-only Data Supplement). Top 5 underexpressed genes included the docking protein 6 (DOK6), calcium/calmodulin-dependent protein kinase type II α chain gene (CAMK2A), myozenin 3 (MYOZ3), an immunoglobulin gene (IGHG4), and tryptophan hydroxylase 1 (TPH1; Table 2). We also found keratin 17 (KRT17), a cytoskeleton protein expressed in skin, but also in blood and brain, among the top 10 genes with underexpression in ruptured versus unruptured aneurysms (Table 2). The overexpression of CD300E (P=0.0039) and MARCO (P=0.0006) in ruptured aneurysm tissue was confirmed in the validation experiment. Comparable to the findings in the RNA sequencing analysis, we again showed underexpression of KRT17 (P=0.06) and CAMK2A (P =0.1085), although for these genes the relative expression differences were not statistically significant in the validation experiment (Table 2; Figure 2). Of the 4 relevant and highly significant genes in ruptured versus unruptured aneurysms, the MARCO gene was also found to be overexpressed in 1 previous study,¹⁰ and the KRT17 gene was also found to be underexpressed in the 2 most recent gene expression studies.^10,12 The other 2 relevant and highly significant genes were not found to be differentially expressed before.

• Download figure
• Download PowerPoint

Functional Network Analysis

Functional analysis of the 1489 differentially expressed genes with an FDR-adjusted P value <0.05 yielded 6 significant Kyoto Encyclopedia of Genes and Genome pathways: lysosome, osteoclast differentiation, Staphylococcus aureus infection, phagosome, leishmaniasis, and Fc γ R-mediated phagocytosis (Table VII in the online-only Data Supplement). After trimming, 306 Gene Ontology terms were identified (Table VIII in the online-only Data Supplement), including many terms involved in immune response, the terms lysosome, lysosome organization, and lysosomal membrane and lumen, and terms involved in cell–cell interaction and in-cell regulation.

Low-Count Gene Analysis

The subanalysis of low-count genes found immunoglobulin κ variable 1D-42 to have >200 counts in unruptured aneurysms and 0 counts in ruptured aneurysms. None of the genes had 0 counts in ruptured aneurysm tissue and >200 counts in unruptured aneurysm tissue.

Discussion

This study found ECM pathways to play a role in aneurysms and pathways involved in immune response in rupture of aneurysms, and identified lysosomes as a new pathway to play a role in rupture. In addition, our results point toward a role for immunoglobulins in the pathogenesis of aneurysms because we found that immunoglobulin κ and heavy chain variable region genes were expressed in aneurysm tissue, but showed no expression in control tissue. One immunoglobulin κ variable region gene was expressed in unruptured and not in ruptured aneurysms. The differences in gene expression found in aneurysms and controls can be the cause, but also be the result of the development of aneurysms. Because intracranial aneurysms are associated with heritable disorders of connective tissue and ECM,¹⁹ changes in the ECM are more likely the cause of the development of aneurysms than its result. As in aneurysms and controls, the differences found in ruptured and unruptured aneurysms can be the cause, but also the result of rupture. Because the time between subarachnoid hemorrhage and clipping was >48 hours in half of our patients, the overexpression of immune response genes in ruptured versus unruptured aneurysms might also be the result of an inflammatory reaction in response to the event of rupture, instead of the cause of rupture. However, 2 previous studies compared biopsies obtained within a range of 2.6 to 24 hours¹⁰ or 6 to 24 hours after rupture²⁰ and those biopsies obtained later showed no differences in gene expression,¹⁰ nor in the degree of inflammatory cell invasion into the wall.²⁰ Furthermore, in a study of ruptured aneurysms from autopsy cases inflammatory cell infiltration was always found to be accompanied by fibrosis, and fibrosis was never present without an inflammatory cell infiltration, even in unruptured aneurysms.²¹ Because fibrosis is considered the end result of chronic inflammatory reactions,²² this strongly suggests that the inflammatory reaction is present before rupture.

Our study identified the expression of both light (ie, of the κ subtype) and heavy chain immunoglobulin genes in aneurysm tissue and its complete absence in control tissue. Two previous immunohistochemistry studies already showed heavy chain immunoglobulins subtypes IgG and IgM to be present in the majority of the investigated aneurysm walls,^23,24 whereas these were only rarely found in control arteries.²³ One of these studies also found sporadic B lymphocytes (which produce immunoglobulins) in unruptured aneurysm tissue, whereas these cells were absent in control arteries.²³ The presence of immunoglobulins and B lymphocytes in the aneurysm wall suggest that the inflammatory reaction in the aneurysm wall, which is not seen in healthy control arteries, is initiated by the humoral immune response, through attraction of inflammatory cells and through complement activation.²⁴ Our study underlines the importance of involvement of genes of the ECM pathway in aneurysms, which was also found in a previous meta-analysis of 5 microarray studies.²⁵ Furthermore, several histopathologic studies have shown degradation of the ECM in intracranial aneurysm tissue.²⁶ We found enrichment of the lysosome pathway in ruptured aneurysms. Lysosomes digest the degradation material from the cell. Phagosomes, another enriched pathway in ruptured aneurysms in our study, fuse with lysosomes after phagocytosis of degradation material. Enrichment of these pathways supports the notion that degradation of the components of the unruptured aneurysm wall is a process leading to rupture,²⁷ but may also be a response to rupture. In a previous genome-wide expression study on blood from patients with aneurysmal subarachnoid hemorrhage taken several years after the subarachnoid hemorrhage compared with blood of healthy controls the lysosome pathway was also found enriched.²⁸ This finding strengthens the idea that the lysosome pathway does not reflect an acute and short-lasting reaction to aneurysm rupture. Pathways found to be involved in rupture in previous studies were inflammation and immune response pathways, ECM degradation, cell adhesion, vascular remodeling, oxidative stress, turbulent bloodflow, proteases, and apoptosis.^8–10,12 Our study also found immune response pathways to be involved in rupture and identified lysosomes as a new pathway. Furthermore, inflammation was a predominant characteristic of ruptured aneurysms in immunohistochemistry studies.²⁶

This study has some limitations. First, aneurysm biopsies could only be taken from aneurysms treated with microneurosurgical techniques, which may induce a selection bias because certain aneurysm characteristics might make aneurysms more suitable for clipping while these characteristics are also associated with rupture.²⁹ Second, as controls we used cortical intracranial arteries obtained from patients with intractable epilepsy. However, we cannot be sure that the composition of these cortical arteries is similar to the composition of the basal arteries of the circle of Willis on which aneurysms arise. Furthermore, we cannot exclude the possibility that the seizures or the epileptogenic focus have altered the cortical vessels, although we found no data supporting such an influence. There are several strengths of this study. First, we did not find larger studies investigating gene expression differences in intracranial aneurysms to date. The large sample size increased power of our study and enabled correction for possible confounders in our analysis, including sex, age, and rupture status of the aneurysm. Furthermore, we used RNA-sequencing methodology, which compared with the previously used microarray technique, has the advantage of requiring less tissue mass as input material (crucial in aneurysm studies because of the small size of the biopsies) and of a genome-wide coverage, enabling the discovery of not yet identified genes involved in the disease. Finally, we used healthy intracranial arteries as controls, which are preferred above extracranial arterial tissue because of the differences in vessel wall composition between intra- and extracranial arteries.

In conclusion, we identified the lysosome pathway as a new pathway for rupture of intracranial aneurysms and found further evidence for the role of the immune response in aneurysmal rupture. Our results also point toward a role for immunoglobulins in the pathogenesis of aneurysms. Our finding that immune response pathways play a role in aneurysm rupture suggest that anti-inflammatory and immune-modifying drugs are interesting candidate therapeutics in the prevention of aneurysm rupture. The presence of immunoglobulins in aneurysm and its absence in control tissue highlights the potential role of immunoglobulin-mediated inhibition of B lymphocytes as an interesting therapeutic intervention in aneurysm development or growth. To identify patients with aneurysms showing signs of increased aneurysm wall inflammation, an imaging technique is required that is able to correctly identify those aneurysms subject to inflammation, for example, by using magnetic resonance imaging with a suitable contrast agent or high resolution vessel wall magnetic resonance imaging.^30,31

Footnotes