RNA sequencing analysis revealed the induction of CCL3 expression in human intracranial aneurysms

Intracranial aneurysm (IA) is a socially important disease as a major cause of subarachnoid hemorrhage. Recent experimental studies mainly using animal models have revealed a crucial role of macrophage-mediated chronic inflammatory responses in its pathogenesis. However, as findings from comprehensive analysis of unruptured human IAs are limited, factors regulating progression and rupture of IAs in humans remain unclear. Using surgically dissected human unruptured IA lesions and control arterial walls, gene expression profiles were obtained by RNA sequence analysis. RNA sequencing analysis was done with read count about 60~100 million which yielded 6~10 billion bases per sample. 79 over-expressed and 329 under-expressed genes in IA lesions were identified. Through Gene Ontology analysis, ‘chemokine activity’, ‘defense response’ and ‘extracellular region’ were picked up as over-represented terms which included CCL3 and CCL4 in common. Among these genes, quantitative RT-PCR analysis using another set of samples reproduced the above result. Finally, increase of CCL3 protein compared with that in control arterial walls was clarified in IA lesions. Findings of the present study again highlight importance of macrophage recruitment via CCL3 in the pathogenesis of IA progression.


Results
Characteristics of human specimen harvested. We collected 55 IA specimen and control arteries from same patients (total 110 specimens) for RNA sequencing or quantitative RT-PCR analyses in the present study during neck clipping of unruptured lesions with written informed consent. Among 55 sample sets collected, 33 sample sets were actually used in the present study (3 for preliminary experiment to purify total RNA, 12 for RNA sequencing analysis, 18 for quantitative RT-PCR analysis). Notably, the average age and also the size of the present cases are closely similar with those reported in a Japanese large cohort (UCAS Japan) 2 (average size; 5.8 mm (all cases), 5.2 mm (only cases subjecting to RNA sequencing analysis), 5.8 mm (only cases subjecting to quantitative RT-PCR analysis) in the present study, 5.7 mm in UCAS Japan, average age; 62.0 year-old (all cases), 62.1 year-old (only cases subjecting to RNA sequencing analysis), 63.1 year-old (only cases subjecting to quantitative RT-PCR analysis) in the present study, 62.5 year-old in UCAS Japan), suggesting that results obtained from the present study represent common pathophysiology of unruptured IAs not the specific population of them.
Gene expression profile in IA lesions by RNA sequencing. Because human IA specimens were harvested during microsurgery, sometimes it took a while before stabilization of these dissected specimens resulting in the degradation of RNA. In addition, as arterial wall is stiff in nature, it is usually technical challenge to purify the adequate quality and quantity of RNA for analyses. Indeed, we purified total RNA from 12 IAs but only 4 of them met the criteria suitable for RNA sequencing analysis. The major reason to exclude 8 total RNA samples from an analysis is an amount of total RNA purified (less than 10 ng) due to degradation. Using these 4 RNA samples from IA lesions and 3 samples from control arteries of same patients (1 samples from the control artery was excluded because of the low yield.), we performed RNA sequencing analyses to obtain gene expression profile in an IA lesion compared with that in a control arterial wall. RIN (RNA Integrity Number) values of these sequenced samples were between 6.6 and 8.1 (Median; 7.5). Read count obtained in the present experiment was about 60~100 million per sample which yielded 6~10 billion bases (read length; 100 base pair). Among these reads, around 25 million fragments per sample were successfully mapped to the reference genome with only the minor contamination of ribosomal RNA (about 0.4~0.9 million) and its value was enough for further analyses.
In the preliminary analysis, majority of changes in gene expression profile in IA lesions compared with that in control arterial walls reflected loss of smooth muscle cells in lesions, which is the well-known histopathological character of IAs, making further analyses difficult. Indeed, about the five times more GO-terms were picked up as down-represented ones in IA lesions compared with that up-represented (Supplementary Tables S1 and S2) and the terms indicating the loss of smooth muscle cells occupied the top (Supplementary Table S1). Therefore, gene expression profile of commercially available primary culture of smooth muscle cells from human carotid artery ( Supplementary Fig. S1) was first obtained and then genes with over 10 TPM (transcripts per million) in these cells were subtracted from the gene list of human IA lesions and control arterial walls obtained from RNA sequencing analysis. Because a primary culture of smooth muscle cells from human intracranial arteries was not available, we alternatively selected a primary culture derived from carotid artery. As the primary culture of smooth muscle cells rapidly proliferated when RNA was purified, this primary culture seems to be in the synthetic state. As a result of these procedures, we identified a couple of differentially expressing genes between IA lesions and control arterial walls. Principal component analysis demonstrated the cluster formation both in gene expression of IA lesions and control arterial walls and cluster of each group occupied completely different location in coordinate axis (Fig. 1), suggesting the presence of notably different gene expression pattern between groups; IA lesions and control arterial walls. We identified 79 over-expressed and 329 under-expressed genes in IA lesions compared with expressions in control arterial walls (Supplementary Tables S3 and S4). Here, under-expressed genes mainly involved smooth muscle cell-related ones, reflecting loss of these cells in IA lesions, and therefore, by gene ontology analysis (GO analysis), GO-terms related with loss of smooth muscle cells such as 'muscle contraction' in BP (biological process), 'contractile fiber' in CC (cellular component) and 'structural constituent of muscle' in MF (molecular function) were indeed picked up as down-representative ones (Supplementary  Tables S5-S7). Notably, only 79 genes were identified as over-expressed ones in IA lesions (Supplementary  Table S3), suggesting the nature of IA as a relatively silent (not active) lesion. Intriguingly, in all three categories, BP (Supplementary Table S8), CC (Supplementary Table S9) and MF (Supplementary Table S10) in GO-term, CCL3, CCL4 and CXCL10 were picked up as over-expressed ones (Table 1 and Supplementary Tables S11 and S12) and all of these genes coded chemoattractants for macrophages. www.nature.com/scientificreports www.nature.com/scientificreports/ Expression of selected genes in another set of human IA specimen by quantitative RT-PCR analysis. Next, to confirm the reproducibility of results from RNA sequencing analysis, quantitative RT-PCR analysis using RNA samples purified from another set of cases (18 cases) was done. In quantitative RT-PCR analysis, we selected CCL3, CCL4, CXCL10 and CLEC5A, a C-type lectin expressed on macrophages, from over-expressed genes in RNA sequencing analysis as targets because the crucial role of macrophage-mediated inflammatory response in the pathogenesis of IA formation has been implicated 6,14,15 in animal studies and consistently the presence of infiltrated macrophages in human lesions are demonstrated [16][17][18] . We found that expressions of CCL3, CCL4 and CLEC5A were significantly over-expressed in IAs compared with those in control arterial walls from same patients consistently with the result in RNA sequencing analysis (CCL3; n = 9, p = 0.0004, CCL4; n = 6, p = 0.002, CLEC5A; n = 9, p = 0.0015) (Fig. 2). As to CXCL10 expression, quantitative RT-PCR analysis failed to reproduce the result in RNA sequencing analysis; the significant increase in IAs compared with control arterial walls. Intriguingly, although contribution of MCP-1-mediated recruitment of macrophages in IA walls to the initiation of experimental IAs in mice has been demonstrated 14,15 , MCP-1 (CCL2) expression tended to be increased in IA walls but differences did not reach statistically significant (n = 10, p = 0.0605) (Fig. 2), suggesting the less contribution of this chemoattractant to IA development in human or the presence of significant diversity in expression level among cases.
Expression of CCL3 protein in human IA lesion. To clarify that over-expression of CCL3 mRNA content in IA lesions compared with control arterial walls indeed reflects protein expression, we immunostained specimens from 2 control arterial walls and 4 unruptured IAs. While signals for CCL3 were seldom detected in 2 control arterial walls, those signals were abundantly detected as scattered ones in whole arterial walls, including medial smooth muscle cells whose arrangement was disturbed and number was remarkably reduced, of all IA lesions examined (Fig. 3, Supplementary Fig. S2), supporting the result from RNA-sequencing and RT-PCR analyses. Intriguingly, the expression of CCL3 protein in IA lesions was observed also in CD68-positive macrophages (Fig. 4), suggesting the formation of self-amplification loop among this type of cells as similarly demonstrated in a mouse model in which CCL2, not CCL3, mediates formation of amplification loop 19 . CCL3 www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
The present experiment revealed the over-expression of genes encoding some chemoattractant factors for macrophages in IA lesions. This finding again highlights the importance of macrophages in the pathogenesis of IAs as indicated by previous studies [14][15][16][17] . The contribution of these chemoattractants to maintenance of the persistent macrophage infiltration in lesions and to the pathogenesis of the disease is thus also suggested. Furthermore, the potential of these chemoattractans as therapeutic targets to treat IAs has been indicated. Intriguingly, however, MCP-1 (CCL2) is not picked up as an over-expressed gene in human IA lesion. Because we and other researchers have demonstrated the crucial contribution of MCP-1 to the initiation and progression of IAs through recruiting macrophages to form inflammatory environment using animal models 14,15 and also some have revealed the presence of this type of cells in human IA lesions as a most abundant inflammatory cell in situ 16,17 , some hypotheses are now raised. First, because both MCP-1 and CCL3 can recruit macrophages in sites of inflammation under inflammatory settings, usage of these chemoattractant proteins among species may be different. Second, in mice model of IAs, because only IAs at the early stage of the disease can be induced, we can only assess the incidence of lesions. Therefore, MCP-1 and CCL3 may play a role at the different stage of IA formation and progression; i.e. MCP-1 regulates macrophage infiltration to induce IA formation at the early stage and CCL3 functions to progress the disease. Third, expression of MCP-1 and CCL3 depends on cytokine milieu in microenvironment of IA lesions because many factors determine expression of these chemoattractants, i.e. transcriptional regulation by NF-κB, c-Ets or so and post-transcriptional stabilization of mRNA 20 . Thereby, abundance of CCL2 may be greatly different among lesions which makes difficult to detect statistical significant difference in expression. Here, intriguingly, we demonstrated in our recent manuscript that NF-κB inhibition specifically in endothelial cells in mice led to almost complete suppression of MCP-1 expression in these cells but infiltration of macrophages in the lesions was still sustained 19 presumably through CCL3 suggesting the redundant role of chemoattractants to recruit macrophages in IA lesions.
Epidemiological study has clarified that only one thirds of IAs in humans ruptures in a lifetime 21 . Further, only a few portion of unruptured IAs annually enlarges and only around 1% of them ruptures 2,22 . These epidemiological findings about natural course of human IAs may imply the nature of unruptured IAs in humans as mostly stable lesions. In line with such an assumption, our RNA sequencing analysis of human unruptured IAs mainly detected changes in gene expression profiles corresponding to loss of medial smooth muscle cells,  www.nature.com/scientificreports www.nature.com/scientificreports/ which is the well-established histological character in IA walls, and only small number of genes was detect as over-expressed ones. Previously reported results from comprehensive gene expression analysis of human IAs have indicated active involvement of inflammatory responses in the pathogenesis of IAs but these studies enrolled mainly ruptured IAs or unruptured ones with a larger size in diameter 8-13 than a typical one. Because rupture itself supposedly evokes intense inflammation presumably due to marked increase of intracranial pressure and excessive activation of sympathetic nerve system, careful interpretation should be done. Indeed, the published report enrolling only unruptured IAs demonstrated the completely different result 23 . Most typically, annotation of the differentially regulated genes in unruptured IAs includes 'antigen processing and presentation pathway' , which is picked up as an over-represented pathway in ruptured IAs, as an under-represented one 23 . The present study may suggest that inflammatory responses in IA walls are transiently enhanced, once after inflammation is triggered by hemodynamic force, followed by the relief of inflammatory responses making IA lesions relatively stable. Macrophages CCL3-dependently infiltrated in lesions thereby may exert anti-inflammatory property and contribute such a relief of inflammation. Based on these assumptions, we hypothesize that rupture of IAs requires the resurgence of inflammation by some unrevealed mechanisms as '2 nd hits' . Potential mechanisms regulating the resurgence of inflammation may be factors previously reported; i.e. thrombus formation [24][25][26] , lipid accumulation 27 presumably by excessive endothelial cell damage or occurrence of aberrant flow condition by morphological changes during the progression of IAs 25,28 . Given the clinical studies in which drugs with anti-inflammatory effect, statins or non-steroidal anti-inflammatory drugs (NSAIDs) [29][30][31] , significantly reduces the risk of rupture of IAs, regardless the trigger of 2 nd hits, inflammation may be crucial also for rupture.
The major limitation about the present study is the small sample size partially because the quality and the amount of total RNA purified from surgically resected specimens were often low. Another limitation is the usage of extra-cranial arteries (middle meningeal artery or superficial temporal artery) as a control. Although the usage of intra-cranial arterial walls where IA is formed of course provides best reference data, we used extra-cranial arteries because the resection of intra-cranial arteries results in ischemic stroke (death of neurons). Also importantly, we could prepare control arteries from the same patients with IA lesions if we used extra-cranial arteries. However, there may be a considerable difference in gene expression profile between intra-cranial arteries, where IAs are formed, and extra-cranial arteries. If present, results obtained in the present study are needed to be carefully interpreted. Also, the subtraction of the comprehensive gene expression to minimize effect of the loss of medial smooth muscle cells using a primary culture of human carotid artery may also interfere with an interpretation of data especially because cultured smooth muscle cells seemto have a synthetic phenotype which enhances expression of some pro-inflammatory genes and thereby may attenuate a contribution of inflammation in lesions. Indeed, some GO-terms were lost after the subtraction (Supplementary Tables S2 and S8). Nonetheless, it may be worthy to note that the importance of macrophages in the pathogenesis is highlighted.  www.nature.com/scientificreports www.nature.com/scientificreports/ aneurysms were included in the present study and other than saccular ones such as fusiform aneurysm, dissecting aneurysm and giant thrombosed aneurysm were excluded. RNA was, then, purified from these samples by using a RNeasy Fibrous Tissue Mini Kit (QIAGEN, Hilden, Germany) according to the manufactures' instructions. After quality check of purified RNA by the RNA analyzer and a Quant-iT RiboGreen RNA Reagent (Molecular Probes, Eugene, OR), the library was prepared using a TruSeq RNA Sample Preparation v2 Kit (Ilumina, San Diego, CA) according to the manufacture's protocol with some modifications. In brief, after purification of polyA + RNA from total RNA samples, the RNA was transcribed into cDNA using a SuperScriptIII instead of SuperScriptII (Thermo Fisher Scientific Inc., Waltham, MA). The 1 st stranded cDNA was subjected into 2 nd stranded cDNA synthesis, followed by end repair reaction and A-tailing reaction. After purification with an AMpure XP (Beckman Coulter, Indianapolis, IN), sequencing adapter provided by a Sample Prep kit was ligated to the double stranded DNA. After purification with an AMpure XP (Beckman Coulter), the amount of the adapter-ligated DNA was then quantified by quantitative PCR analysis using KAPA Library Preparation Kits with a Real-Time Amp (Kapa Biosystems Inc., Woburn, MA) to estimate the optimal PCR cycle number for library amplification. The amplification of library was performed using KAPA Library Amplification Kits (Kapa Biosystems Inc.) with optimized PCR cycles. The gene expression profile was then obtained by RNA sequencing (100 bp paired-end read sequencing, HiSeq2500, Illumina). The criteria for RNA quality to perform RNA sequencing analysis are that a sample amount is over 100 ng and the ratio of A260/280 and A260/A230 is both over 1.8 (http://www2.clst.riken. jp/genas/services/rnaseq.html). The raw data was processed by using the MOIRAI workflow system (PMID: 24884663) 32 , where bcl2fastq (version 1.8.4) was used for basecalling and demultiplexing, FASTQC (version 0.11.3) for sequence quality check, fastx_clipper (version 0.0.13.2) for adaptor trimming and tophat (version 2.0.12) for alignment with the human reference genome (hg19). The mapped reads were counted per gene by HT-seq (version 0.6.1 p1, PMID: 25260700). Normalization of gene expression profiles and differential analysis were conducted by using edgeR (version 3.8.6, PMID: 19910308) with false discovery rate 0.01 and two fold change as a threshold for statistical significance, after genes expressed in primary culture of smooth muscle cells from human carotid artery were excluded.

Methods
Data from RNA sequencing analysis is deposited to the BioProject database at NCBI (http://www.ncbi.nlm. nih.gov/bioproject) as ID number #PRJNA553307.
primary Culture of Vascular smooth Muscle Cells from human carotid artery. The primary cul-