Blood Transcriptome Analysis of Septic Patients Reveals a Long Non-Coding Alu-RNA in the Complement C5a Receptor 1 Gene

Many severe inflammation conditions are complement-dependent with the complement component C5a-C5aR1 axis as an important driver. At the RNA level, the blood transcriptome undergoes programmed expression of coding and long non-coding RNAs to combat invading microorganisms. Understanding the expression of long non-coding RNAs containing Alu elements in inflammation is important for reconstructing cell fate trajectories leading to severe disease. We have assembled a pipeline for computation mining of new Alu-containing long non-coding RNAs by intersecting immune genes with known Alu coordinates in the human genome. By applying the pipeline to patient bulk RNA-seq data with sepsis, we found immune genes containing 48 Alu insertion as robust candidates for further study. Interestingly, 1 of the 48 candidates was located within the complement system receptor gene C5aR1 and holds promise as a target for RNA therapeutics.


Introduction
Alu elements are short interspersed transposable elements (SINEs) that have propagated in the primate lineage with more than one million copies and constitute approximately 10% of the human genome. Alu elements are dispersed throughout the genome with a bias toward gene-rich regions, concentrated in introns and untranslated regions (UTRs) of genes. Evolution has given rise to Alu subfamilies where AluJ and AluS represent the earliest evolved, and AluY the most recently evolved [1,2]. Alu elements are non-autonomous retrotransposons but may still transpose by hijacking enzymes expressed from genomic long interspersed transposable elements (LINEs) [3].
Alu elements are derived from the 7SL RNA and are therefore originally transcribed by RNA polymerase III (Pol III). As Alu elements have propagated through the primate genome, more than half of all known Alu elements are located within introns and UTRs of Pol II-transcribed genes, and it can be challenging to determine whether an Alu element is actively transcribed as an independent primary transcript (Pol III) or processed from a mRNA or a long non-coding RNA (lncRNA) Pol II transcript [4,5]. Pol III-transcribed Alu elements embedded within a Pol II transcriptional unit possess their own transcriptional start site (TSS) and regulatory mechanisms. Although transcriptional initiation can be immediate genomic neighbouring genes [31,32]. Thus, the aim of the present study was to systematically evaluate~3500 canonical immune genes in order to predict and identify new putative Alu-containing lncRNAs in sepsis. A total of 48 Alu insertions were found in 26 immune genes, and by filtering for independent transcriptional start sites, 20 strong candidates were identified. By these criteria, we found the C5aR1 receptor gene to contain a new Alu-lncRNA, which was further validated with RNA-seq data from the whole blood model.

Materials and Methods
We assembled an "in house" pipeline based on published genomic tools and databases for the identification of transcribed Alu-lncRNAs in immune gene coordinates. An overview of the Alu-lncRNA pipeline for setting the scope of the intersection between the gene and Alu elements is presented in Figure 1.
in the study, 88 were used for downstream transcriptome analysis, 40 from healthy controls, and 48 from patients diagnosed with sepsis, septic shock, or severe infection. The findings were confirmed by a second sepsis cohort and our in-house RNA-seq data from an ex vivo whole blood model, where whole blood was activated with Staphylococcus aureus. Previous studies have shown that many intragenic and antisense lncRNAs regulate their immediate genomic neighbouring genes [31,32]. Thus, the aim of the present study was to systematically evaluate ~3500 canonical immune genes in order to predict and identify new putative Alu-containing lncRNAs in sepsis. A total of 48 Alu insertions were found in 26 immune genes, and by filtering for independent transcriptional start sites, 20 strong candidates were identified. By these criteria, we found the C5aR1 receptor gene to contain a new Alu-lncRNA, which was further validated with RNA-seq data from the whole blood model.

Materials and Methods
We assembled an "in house" pipeline based on published genomic tools and databases for the identification of transcribed Alu-lncRNAs in immune gene coordinates. An overview of the Alu-lncRNA pipeline for setting the scope of the intersection between the gene and Alu elements is presented in Figure 1.

Whole Blood Model Induced with S. aureus
Whole blood from 6 healthy donors was collected after informed consent according to the guidelines from the regional ethics committee. All equipment, tips, and solutions were RNase and endotoxin-free. Fresh human whole blood was drawn into a PAXgene Blood RNA Tube (PreAnalytix, Hombrechtikon, Switzerland) as a baseline sample [35].
In addition, blood was drawn into 9 polypropylene tubes, 4.5 mL (Nunc Cryo Tubes, Thermo Fisher Scientific, Waltham, MA, USA), containing lepirudin (Refludan ® , European Medicines Agency, Amsterdam, The Netherlands), final concentration 50 µg/mL. Aliquots of the lepirudin blood were divided into 4.5 mL Nunc cryotubes supplemented with phosphate-buffered saline (PBS) containing calcium and magnesium (Sigma Merck Life Science AS, Darmstadt, Germany) and preheated in a block heater at 37 • C for 5 min. Heatinactivated S. aureus (1 × 10 8 cells per mL) strain Cowan (ATCC 12598) or PBS was then added. Samples were incubated on a Rock'n roller at 37 • C for 0, 30, 60, and 120 min. The inflammatory reaction was stopped by the addition of EDTA (10 mM final concentration). Samples were centrifuged at 1500× g for 15 min at 4 • C. Plasma was removed and frozen at −80 • C. The plasma volume was replaced by PBS, and the blood tubes were briefly vortexed before adding a volume of Paxgene solution (2.73 mL per mL blood). Samples were stored at −80 • C for RNA isolation.

RNA Isolation for RNA-Seq
MagMAX™ for Stabilized Blood tubes RNA Isolation Kit (Ambion by Life Technologies, Carlsbad, CA, USA) was used for RNA extraction. The concentration and purity of isolated total RNA was measured using a NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). RNA integrity number (RIN) was assessed by microfluidic capillary electrophoresis using an Agilent 2100 Bioanalyzer and the RNA 6000 Nano Chip Kit (Agilent Technologies, Santa Clara, CA, USA) RNA samples exhibited RNA integrity numbers (RIN) between 7.8 and 9.6 (mean RIN 8.8), indicating optimal RNA quality for downstream applications. RNA from the 6 donors was pooled based on concentration for RNA-seq and kept separate for other downstream experiments at −80 • C.

Illumina Sequencing
Next-generation sequencing data were prepared at Eurofins Genomics GmbH (Ebersberg, Germany) using commercially available chemistry. Total RNA was used as starting material; prior to library preparation, the rRNA was depleted using Ribo-Zero™ Magnetic Kit (Human/Mouse/Rat) (Illumina, San Diego, CA, USA). Subsequently, the library preparation was conducted using chemistry from New England Biolabs (NEBNext, Ipswich, MA, USA). All libraries were pooled and sequenced across 3 lanes on a HiSeq 2000 instrument with chemistry v3. The sequencing was performed using the original chemistry provided by Illumina with 100 bp paired-end reads. The sequencing data were included in this study to verify the whole blood model as an in vitro imitation of severe blood infections. The sequencing data are only presented here in Figure 3 and Table S4. This sequencing data is available on request.

Ion S5 Stranded RNA Sequencing
For the Ion S5 sequencing, the whole blood model was followed as described for the Illumina sequencing. One µg of total RNA was subjected to rRNA depletion using the Low Input RiboMinus™ Eukaryoute System v2 (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA). RNA libraries were constructed using Ion Total RNA-seq Kit v2 (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's protocol. Template preparation of libraries was carried out on the Ion Chef™ Instrument (Thermo Fisher Scientific, Waltham, MA, USA) and sequenced on Ion 540 chips using the Ion GeneStudio™ S5 System (Thermo Fisher Scientific, Waltham, MA, USA). The sequencing data were applied to this study in order to verify the transcriptional direction of the Alu-RNAs in Table 1 and are presented in Figure 3. The sequencing data are available on request.
The bam mapping files were sorted with samtools version 1.12 (htslib.org).

Alu Coordinates Intersection with Genes for Expression Analysis
For expression analysis of Alu insertion, we first downloaded the Repeatmasker track from the UCSC genome browser using the table browser tool. https://genome.ucsc.edu/ (accessed on 1 December 2021).

Counting Alu Transcript Levels with the Homer Software
The Homer software (v4.11) [39] (http://homer.ucsd.edu/homer/) (accessed on 1 December 2021) is easy and configurable for implementing your own coordinates and reporting several statistical tests and normalisation of the data, such as VST (variance stabilization), Deseq, and raw reads. First, we performed the script makeTagDirectory to count all the reads for each sample bam file; we ran the following parameters to obtain transcriptome count data: analyzeRepeats.pl Alu_genes.gtf hg38 -strand both -min 400 -raw -dfile reads_to_analyze > out.csv.
The out.csv was fed into Strand NGS software, ver 4. The conditions were set according to [30]. First, we carried out (uniform manifold approximation and projection) UMAP on all 1173 transcripts with the following parameters: The list of 1173 insertions were intersected with the Ensembl reference to represent 716 genes.
Filtering for robust and focused Alu-lncRNA candidates: in the Strand software, we filtered the results by a fold change of 1.3 and raw reads set to 1000 in one of the samples. This generated a list of 48 candidates that were manually inspected in IGV version 2.11 with the corresponding bam files. https://software.broadinstitute.org/software/igv (accessed on 1 December 2021).
All p-values were calculated by a two-way t-test and adjusted for false discovery rate by the Benjamini-Hochberg correction. Significant p-values were defined as <0.001.

Alu Insertions in the Human Genome
Based on the dataset and software used in this study, we identified a total of 1.2 million Alu element insertions (UCSC) in the human genome (Hg38). These elements were divided into three families: AluJ (320,000 insertions), AluS (727,000 insertions), and AluY (149,000 insertions). The focus of this study was to investigate the intragenic Alu elements, and we found approximately 800,000 Alu insertions located within genes (coding and non-coding). We use the term "gene" as defined by ensembl.org-built from gene-wise alignments of the human proteome and alignments of human cDNAs. We analysed the positions of Alu elements within the gene bodies ( Figure 1) and found that the most frequent Alu insertion sites were in introns (close to 90%, Figure S1). Less than 5% of Alu insertions were located in the 3 UTR. The subfamilies AluJ, AluSx, and AluSx1 were the most common Alu insertions observed ( Figure S1b).

Alu-RNAs in Blood from Sepsis Patients and Healthy Controls
Expressed Alu elements show high tissue specificity [5], and Alu-RNAs have been reported to be involved in cell differentiation and in responses to cellular stress, such as infection. We investigated the Alu element expression profile in the transcriptome of patients with severe inflammation. Transcriptome data from 12 patients with infection, 18 patients with sepsis, 18 patients with septic shock, and 40 healthy controls from the RNA-seq analysis of blood from sepsis patients and healthy controls-study (PRJNA647880) [30]-were included in this study (Table S1). The processed RNA-Seq data were filtered based on the following criteria: (1) Transcripts had to contain Alu elements in intragenic locations (excluding exons), and (2) Table S2). We found that the Alu-containing transcripts were clustered based on health conditions, where the healthy controls and patients with inflammation constitute separate branches. About half of the transcripts showed an increased expression in healthy controls versus participants with inflammation, while a decrease in expression was observed in the other half. Although the hierarchical clustering separates healthy versus inflammation, the analysis did not differentiate between the inflammatory condition infection, sepsis, and septic shock. UMAP (Figure 2b), on the other hand, supported the hierarchical clustering, but in addition, it shows that the dataset was separated between all four conditions. Normalised expression values are also shown as box plots ( Figure 2c); interestingly, the most extreme values were found in septic shock, the most severe form of sepsis.
The Alu-containing transcripts with the highest expression (>1000 raw reads in one or more samples) and highest fold change between patients with inflammation and healthy controls (fold change threshold 1.3) were considered the most interesting candidates with regard to a potential regulatory role. These criteria were met in 58 genes that had at least one Alu insertion (Figure 2d), and 26 of these (45%) were considered immune genes by the Innatedb.com database (accessed on 1 December 2021) (indicated by red letters in Figure 2d). Although not defined as immune genes by this definition, the remaining genes (black letters in Figure 2d) were to a large extent involved in transcriptional regulation, potentially leading to genome-wide changes in transcription and cell differentiation. A closer examination of the 48 Alu elements embedded in the immune genes from Figure 2d is presented in Table 1. Here we assessed genomic features such as location within the gene, Alu subfamily, adjacent TSS, and orientation of the Alu element relative to the gene in which they are inserted. TSS positions were of interest in order to identify standalone ncRNA transcripts. The TSS coordinates used in this study originate from CAGE data (Cap Analysis Gene Expression) from the FANTOM5 project. The TSS data were captured and aggregated from over 400 different human cell types using the CAGE protocol. In general, the CAGE protocol captures and links the 5 CAP structure of a cells transcript, and by adding a linker to the tagged 3 end, RNA-seq can be used to generate and map transcriptional start sites in the sample [41]. The p-values for the 1173 differentially expressed normalised Alu transcripts are given in Table S2. All differentially expressed normalised Alu transcripts in Table 1 were highly significant (p ≤ 0.001).   Figure 1a, but in addition, shows that the dataset is separated between the conditions. (c) Boxplot showing normalised expression transcribed from intragenic regions and containing Alu elements. The boxes mark the medians and the upper and lower quartiles, and the whiskers extend to the most extreme values. (d) Heatmap of Alu-containing transcripts with the highest expression (>1000 raw reads in one or more samples) and highest fold change (threshold 1.3) between patients with inflammation and healthy controls. Alu transcripts from within the same genes are aggregated and produced transcripts from 58 genes that had at least one Alu insertion. Genes in red letters represent immune genes as defined by the Innatedb.com database (accessed on 1 December 2021). Box connected to AluSx1 C5aR1 shows p-value of p ≤ 0.001.
To further validate the robustness of our results, we applied a second RNA-seq dataset from peripheral blood mononuclear cells (PBMC) of sepsis patients, including healthy controls at Taizhou Hospital (PRJNA607653) (Table S1), which was analysed through the same pipeline ( Figure S3, based on Table S3). Reproducing the biological results in different studies increases its validity and reliability of the data. We found a good correlation between the highly transcribed, differentially expressed Alu-RNAs. Of the 26 immune genes in Figure 2d, 15 genes were also included in the corresponding figure from the Taizhou dataset ( Figure S3b).
We also introduced our own Illumina and Ion Torrent RNA-seq data from the whole blood model. This is a relatively uncomplicated in vitro model where we can measure how whole blood reacts to different stressors such as bacterial infection. We incubated blood from healthy donors with Staphylococcus aureus for two hours to mimic a severe bacterial infection. By obtaining similar data from the whole blood transcriptomes of sepsis patients and the bacteria induced ex vivo whole blood model, we strengthened the value of the model. Applying different sequencing platforms may illuminate variety in consistency between similar studies due to bias in the technologies. Illumina transcriptome data from the whole blood model were extracted based on a >200 reads threshold and intersected with Alu-RNA data from Table 1, obtaining transcripts from 16 Alu elements, nested in immune genes (Table S4). These results show that the in vitro whole blood model can to some extent mimic the expression in patient data.
Transcriptome data from the whole blood model sequenced by Ion Torrent were applied and manually inspected to verify the direction of the transcripts. All transcripts listed in Table 1 have the same orientation as the gene in which they are embedded. Interestingly, more than half of the transcribed Alu elements presented in Table 1 were found in the 3 UTR (27 in 3 UTR and 21 in introns), and most of the Alu elements with an associated TSS were also from a 3 UTR location. None of the Alu RNAs from Table 1 were located in the 5 UTRs. The differential expression and association with TSS suggested that these transcripts were 3 UTR-derived ncRNAs with regulatory functions in cis or trans and which have become less dependent on the processing of the host gene.

A Differentially Expressed Alu-Containing lncRNA Is Transcribed from the C5aR1 3 UTR
One interesting regulatory Alu element candidate was found in the 3 UTR of the complement component 5a receptor 1 (C5aR1) gene, also called cluster of differentiation 88 (CD88) gene. C5aR1 is a central 7-span G-coupled receptor in the complement cascade. Activated by the anaphylatoxin C5a, this receptor has a key role in the initiation and maintenance of several inflammatory responses in neutrophils and monocytes. We found that AluSx1 located in the 3 UTR of the C5aR1 gene was highly and differentially expressed during inflammation compared with healthy controls (Figure 2d, right box and Table 1) and that this 3 UTR has an associated TSS located 300 nucleotides upstream of the Alu element ( Figure 3, 3 UTR TSS marked as TSS*). This genomic feature has the potential of creating an Alu element containing lncRNA of about 1000 nucleotides (sequence in Figure S2). This finding was further supported by blasting the sequence in the NCBI EST database. We found two full-length 5 enriched EST transcripts with overlapping sequences (AL571812.3 and AL572873.3) and several EST transcripts that cover the 3 part of the C5aR1-3 UTR-lncRNA including the Alu element (CA309875.1 and BP297702.1).
The AluSx1 in the C5aR1 3 UTR was also found to be upregulated in the Taizhou study ( Figure S3, Table S3) and our Illumina and Ion Torrent RNA seq data from the whole blood model. When we compared the C5aR1 3 UTR from the sepsis studies and the in vitro model; we observed an increase in transcription when comparing healthy control raw reads versus septic shock and when comparing raw reads from the whole blood model activated with bacteria versus healthy control. To investigate the independence of the Alu-lncRNA, we examined the RNA-Seq read mapping in the healthy and septic patients. In both cases, we saw a drop in read map alignment at the identified TSS (TSS*, Figure 3, representative patients included in the figure), which can give evidence towards the Alu-lncRNA as an independent transcript. These results were further supported by the Taizhou dataset ( Figure S3). study ( Figure S3, Table S3) and our Illumina and Ion Torrent RNA seq data from the blood model. When we compared the C5aR1 3′UTR from the sepsis studies and vitro model; we observed an increase in transcription when comparing healthy c raw reads versus septic shock and when comparing raw reads from the whole model activated with bacteria versus healthy control. To investigate the independe the Alu-lncRNA, we examined the RNA-Seq read mapping in the healthy and sep tients. In both cases, we saw a drop in read map alignment at the identified TSS Figure 3, representative patients included in the figure), which can give evidence to the Alu-lncRNA as an independent transcript. These results were further support the Taizhou dataset ( Figure S3).

Discussion
Delineation of the transcriptome landscape in septicaemia and severe inflammation is crucial for understanding the inflammation process in key primary immune cells such as neutrophils and monocytes. Whole blood as a tissue type is easily available for diagnostics and research, and molecular signatures of disease states can be discovered and assessed by proteomics and transcriptomics. In the current study, we report regulatory RNAs in the 3 UTR of key genes involved in inflammation that contain Alu elements with robust expression patterns.
By applying a strict cut-off in the analysis, regarding read abundance and differential expression, we found robust Alu-RNA candidates belonging to the AluS, AluJ, and AluY families spanning intronic regions (n = 21) and 3 UTRs (n = 27). Both inverted and sense Alu elements (relative to nested gene and transcription) were observed. All expressed Alu elements were transcribed in the same orientation as the nested host gene. Eighteen of the Alu transcripts had a TSS located upstream (<1000 nucleotides) of the transcribed Alu element, while two Alu transcripts had a TSS in the 5 part of the Alu element. From this total of twenty transcripts with associated TSSs, fourteen were from the 3 UTRs, and the remaining six were from intronic coordinates of immune genes. These findings were supported by addition of a second patient dataset (PRJNA607653) and the whole blood model using Illumina sequencing for paired-end and Ion Torrent for directional verification of transcripts. Differences in the datasets can be both biological and technical. RNA was isolated from patient whole blood, mononuclear cells, or whole blood drawn from healthy donors which were stimulated with bacteria in vitro. Different sequencing platforms with divergent library preparations will also affect the overall results. All these factors will influence the output of the analyses. We found a considerable reproducibility between the different datasets.
The distribution of Alu subfamilies is known to vary considerably across different transcript types. Kim and co-workers reported that AluJ subfamilies were overrepresented in lncRNAs, while AluS was more common in mRNA transcripts [8]. Furthermore, Su and co-workers put forward the idea that Alu elements may have evolved toward an enhancer function in the human genome, which were exemplified by old Alu subfamilies such as AluSx, AluJo, and AluJb [24]. These families were shown to have gained regulatory motifs, such as transcription factor binding sites and were enriched in histone post-translational modifications. We argue that Alu-containing transcripts in the immune genes represent regulatory RNA where the Alu element has gained a biological role. We found these Alu elements mainly from the subfamily AluS (n = 35, Table 1), and particularly from AluSx. This is consistent with the observation that AluS is overrepresented in mRNA transcripts. The role of AluSx as a transcriptional regulator is also reported by others. Cantarella and co-workers [42] showed that overexpression of AluSx in human fibroblast cell lines resulted in highly significant enrichment of pathways involved in cell cycle progression. This indicates a role as a transcriptional regulator of cell cycle genes. The regulatory role is also closely linked to the structural properties of Alu-RNA. Alu-RNAs have a defined folding, and the structure is key in interaction with proteins, DNA, and other RNAs. The structural importance of Alu-RNA, including AluSx, in regulation is emphasized in the study by Aune and co-workers [43], where the ability to activate innate immune responses resides on structural changes by A-to-I editing.
Several studies have investigated the differential expression of Alu-RNA in the brain during disease versus healthy state. Cheng et al. [44] showed how Alu-RNA regulated gene expression by binding RNA polymerase II and suppressing transcription in normal tissue. During stress, expression of Alu-RNA was upregulated, causing genome-wide transcriptional repression. This capacity to suppress RNA polymerase II is again regulated by Alu-RNA self-cleaving activity, generating another layer of regulation. This Alu RNA processing is increased in Alzheimer's disease. Another study on Alzheimer's disease found that the long non-coding RNA BC200 (BCYRN1) includes a left monomer Alu element. During normal aging the BC200 RNA is reduced, while it is significantly upregulated in brain tissue of patients suffering from Alzheimer's disease [45]. In brain tumours, the opposite effect is reported; Hwang and co-workers compared Alu-RNA expression between glioma brain tumour and normal tissues. They found downregulation of Alu-RNA in almost all pathologies of gliomas [46]. Common across all three studies was a change in expression between tissue from disease and healthy controls. The regulatory role of Alu-RNA is emphasized by the deregulation in disease.
Alu elements located in the 3 UTR of mRNA genes are known to be enriched in distinct cellular functions [47]. Liang and Yeh found that antisense Alu elements were common in 3 UTRs of gene transcripts associated with neurological and developmental processes, while sense transcripts were enriched in genes involved in immunological processes [44]. However, our observations of differentially expressed Alu-RNA embedded in immune genes (Table 1) appear to deviate from the findings by Liang and Yeh [44]. Most of our Alu-containing 3'UTRs have an associated TSS, and we favour the idea that Alu-RNAs from 3 UTR transcripts are, at least partly, independent of their host mRNA transcripts. Interestingly, we found that the AluSx1 element in C5aR1 3 UTR had a clear TSS and indicates the formation of an Alu-lncRNA. Mapping of all datasets, both in vivo and in vitro, shows a scarcity of transcripts at the location of the TSS, suggestive of a separate 3 UTR transcript. The AluSx1 element was highly and differentially expressed during inflammation compared with healthy conditions. The C5aR1 sits in a central hub of the complement inflammation cascade by directly binding C5a to induce inflammation in sepsis [35]. A major limitation of our work is the lack of functional studies validating the role of the AluSx1 element as a transcriptional regulator. Such studies, including siRNA-mediated knockdown and luciferase reporter assays coupled with structural analysis, should be pursued in the future. However, the role of this group of Alu-containing ncRNA appears intriguing. They are differentially transcribed in immune cells during inflammation from an immune gene body, which suggests that these transcripts regulate either immunological or cell differentiation processes. Their role may include cis-regulation of the nested gene.
Here, a mechanism such as transcription or translation initiation or repression, by binding DNA at promoter or enhancer sites, recruitment of regulatory proteins, or binding the mRNA directly, may be feasible.
Alu elements appear common within immune-related genes. In the search for new candidates for diagnostics and possible targets for RNA drugs such as gapmers and other RNA-mers [48], we found new transcripts containing Alu elements linked to canonical immune gene expression. Of particular interest is a new Alu element, AluSx1 element, in the 3 UTR of the complement component 5a receptor 1 (C5aR1) gene. Our study reveals the C5aR1 3 UTR as a promising target for studying the immune response in severe inflammation.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: The study was granted permission to use blood from healthy individuals after informed consent following Regional Ethics committee IDs: P REK NORD 32/2004 and 2013/1801/REK nord. Data Availability Statement: WBM S5 and Illumina data are available on request to the corresponding author: bard.ove.karlsen@nlsh.no.

Conflicts of Interest:
The authors declare no conflict of interest.