Data report on gene expression after hepatic portal vein ligation (PVL) in rats

Due to the increasing life expectancy and life style, the incidence of primary liver cancer is steadily rising. Worldwide, it is the fourth leading reason of cancer-associated death [29,33]. The etiology of liver cancer is highly diverse including besides others viral, toxic, nutritional, etc. risk factors that render treatment options as complex as different pathogenic pathways are involved [22,13,12]. Besides novel pharmacological and cell therapy approaches, surgical interventions are the only potentially curative strategies [3]. Among these, hepatic resection intends to remove the solid tumor taking into account tumor location and size as well as vascular supply of the parts of the liver to be removed and of the remaining liver. Since the lobar organisation of the liver is mirrored by separate venous blood drainage of the lobes, liver resection in general means removal of the tumor bearing lobe(s). The loss of even more than 60 % of the liver mass may be tolerated given that the future liver remnant provides sufficient post-surgery regenerative and metabolic function. To enhance the function of the future liver remnant, the technique of portal vein embolisation is used clinically. This procedure aims at increasing the future liver volume anticipating that volume equals function, i.e., volume growth of the non-ligated lobe(s) compensates for the surgical liver mass loss. However, volume does not necessarily reflect function. Therefore, in order to assess post-surgery hepatic metabolic and regenerative capacities, it is necessary to characterise changes in gene expression in the atrophic and hypertrophic lobes, respectively, and to correlate these changes with the prospective functional efficiency of the liver remnant after resection of the tumor bearing ligated liver lobe(s).Experimentally, portal vein ligation is comparable with portal vein embolisation applied in clinical settings. For the data collection presented here, we applied a model of 60% portal vein ligation in the rat, i.e., the left median and the lateral left lobes as well as the right superior and inferior lobes were deprived by ligation of the portal vein, while the right median and the left superior and inferior lobes remained unaffected (see Fig. 1), accordingly modified as described [31,25].Although non-coding RNAs (ncRNAs) such as microRNA (miRNA) or long ncRNA (lncRNA) have been shown to be deeply involved in the pathophysiology of almost all acute and chronic liver diseases, many of the especially liver-specific ncRNAs, have been not even annotated yet [24]. Our group developed new transcriptome assembly pipelines by combining existing tools to identify the different isoforms of mRNAs and lncRNAs [16,11].Non-coding RNAs in human genome. From the pilot project of ENCODE we know that less than 3 % of the human genome code for proteins [10,9]. The remaining genome is divided into 45 % repetitive elements (SINEs, LINEs, transposons), 26 % introns and other unique non-coding DNA [15]. The question of their meaning has raised and is only being answered slowly. By now, we know that at least 80% of the human genome has a function [9]. Some ncRNAs are known to be located within introns, in 5' and 3' untranslated regions (UTRs), antisense to protein-coding sequences or just close to them. However, the in silico identification of the ncRNAs is still a huge challenge. Due to their diverse and fast evolving sequence they are identified most efficiently by a combined in silico/ex vivo approach, that is sequencing the transcriptome and establishing tissue/organ/organism-specific bioinformatical tools. During the last decade, miRNAs, as regulators of various cell processes, received major attention. However, currently there are 3016 ncRNA families described (Rfam v.14.1 [17]), of which 800 are associated to human, covering more than 16.000 genomic regions 1 . Additionally, the existence of long non-coding RNAs (>200 nt) containing introns themselves is estimated in humans by GENCODE v.33 to 17952 loci2 . Our knowledge about lncRNAs is limited and no general computer program for their identification including secondary structure information and protein interactions is developed, yet.Non-coding RNAs in liver. Although only a fraction of liver-specific ncRNAs are known, here we summarize important examples of ncRNA participating in the pathogenesis of different forms of liver disease and how they can be used as therapeutic tools or targets for novel treatment paradigms, following the suggestions of Roy et al. [24], see Tab. 1. A large number of ncRNA genes being involved in metabolic processes, inflammation and immune response are differentially expressed during aging and other biological processess. We therefore expect to see ncRNAs and especially miRNAs to play a role in senescence and inflammation in rat liver when comparing PVL with healthy liver [5].MicroRNAs in liver. MicroRNAs (miRNAs) are a class of small, non-protein coding RNAs that play a crucial role in mediating post-transcriptional gene silencing [6]. These molecules, typically 18 to 25 nucleotides in length, act as regulators of gene expression by binding to the 3' untranslated region of target genes [4]. Interestingly, the intricate machinery governing miRNAs holds significant importance not only in the context of liver diseases [30], but also in processes related to portal vein ligation [27].In the realm of liver biology, miRNAs exert profound effects on various aspects of liver function and pathology. They participate in regulating processes such as hepatocyte proliferation [27], differentiation, apoptosis, and lipid metabolism [1]. Dysregulation of miRNAs has been implicated in the pathogenesis of liver diseases ranging from viral hepatitis [26] to hepatocellular carcinoma [23]. Moreover, the role of miRNAs in portal vein ligation, a surgical procedure often used in experimental models to study liver regeneration, underscores their significance beyond disease states [28]. MiRNAs are intricately involved in the molecular pathways underlying the response to portal vein ligation, influencing the regeneration capacity of the liver and impacting overall hepatic function [8]. Overall, the multifaceted involvement of miRNAs in liver biology and portal vein ligation highlights their versatility and potential as therapeutic targets in the context of liver diseases and surgical interventions.Experiments were conducted from March 1 until April 13, 2022. Male Sprague-Dawley rats (Charles river, 320-440 g) were housed under a 12 h dark/light cycle at ambient temperature with free access to food and water. The experiments were run with four different groups including four animals each for the PVL and the sham operation at two different time points (2 and 5 days) after surgery (16 animals). An additional weight-matched two animals without any treatment were run along as control in order to identify potential surgery-induced differences as compared with the sham-operated animals. The two time points were chosen following the rational that post-surgery regeneration after PVL might follow similar kinetics as regeneration after partial hepatecomy featuring a maximum regenerative response after 2 days and returning to starting conditions again after 5 days [2,14]. The PVL operation was a modified procedure as described previously [32] by ligating in addition to the left median and left lateral portal vein before the bifurcation, the right superior and inferior portal veins before the bifurcation. In sham-treated animals, the abdominal cavity was opened by a midline incision and closed again thereafter. After the sham or the PVL operation, the animals were left under housing conditions. At 2 and 5 days after surgery, respectively, the animals were euthanized under 2,% isoflurane anesthesia and livers explanted. The weight-matched control animals without any treatment were run along. Tissue pieces (ca. 50 mg) from the left (LML, ligated)and the right (RML, non-ligated) median lobes (Fig. 1) were immediately collected in 750 µl QIAzol reagent (Quiagen GmbH, Hilden, Germany) in 2 ml reaction tubes (Biozym Scientific GmbH, Hessisch Oldendorf, Germany) containing 5 PreCelly beads (PEQLAB Biotechnologie GmbH, Erlangen, Germany), snap frozen and stored at -80°C until further analysis. In total, 52 samples from 18 animals were passed to RNA extraction and subsequent sequencing. Left: During PVL surgery two portal vein branches were ligated (blue lines). Samples were taken from the right median lobe (RML, non-ligated during PVL) and the left median lobe (LML, ligated during PVL) for all animals. Right: PVL and sham surgery were performed for eight animals each. Half of the animals were sacrificed after two days (2d), the other half after five days (5d). Control animals were not operated at all. For PVL and sham animals, total RNA libraries were made, for PVL and control animals, small RNA libraries, resulting in 32 and 20 samples for total and small RNA, respectively.Frozen samples were thawed and centrifuged twice for 10 sec at 5.500 rpm in the PRECELLYS ® 24 homogenizer (VWR International GmbH, Darmstadt; Germany), and mixed with 150 µL chloroform for 30 sec using a vortexer. After incubation for 10 min on crashed ice, phases were separated by centrifugation at 4 °C for 15 min and 11.000 rpm in the Fresco 21 centrifuge (Haereus, Hanau, Germany). The aqueous phase was collected, the RNA precipitated with 500 µL isopropanol overnight at -20°C and subsequently centrifuged for 10 min at 4 °C in the benchtop microcentrifuge. The pellet was washed with 500 µL 3 M sodium acetate and the suspension pelleted for 10 min at 4 °C and 11.000 rpm. This step was repeated, the final pellet dissolved in 300 µL RNA grade water at 4 °C for 2 h. Thereafter, the RNA was precipitated with 600 µL ethanol abs. at -80°C for 1 h and pelleted for 10 min at 4 °C and 11.000 rpm. The pellet was washed with 500 µL 80% ethanol and the pellet collected for 10 min at 4 °C and 11.000 rpm. After repetition of this step, the final pellet was collected and shortly air dried. The RNA was dissolved in 75-100 µL of RNA grade water overnight in the fridge. The RNA content and quality were determined using the Amersham NanoVue equipment (GE Healthcare, Freiburg, Germany).Total cellular RNA from the tissues were isolated. cDNA libraries were prepared for all 16 operated animals, utilizing the NEBNext Ultra II Directional RNA library preparation kit including rRNA depletion (rRNA-) according manufactures instructions. The complete experiment was performed in four independent, biological replicates; thus, 32 rRNA-libraries were sequenced on a NovaSeq6000 run yielding 251 bp reads on paired-end mode (further referred to as "total RNA dataset"). For all animals which underwent a PVL surgery, additional cDNA libraries were prepared utilizing the TruSeq Small RNA library preparation kit according manufactures instructions, to sequence especially small RNAs. Additionally, for the two control animals, cDNA libraries were prepared utilizing the TruSeq Small RNA library preparation kit according manufactures instructions. The resulting 20 libraries were sequenced on a NovaSeq6000 run yielding 51 bp reads on paired-end mode (further referred to as "small RNA dataset").The sequencing data was quality controlled using fastQC (v0.11.9) 3 . Adapter sequences and low quality bases were removed by trimmimg with fastp (v0.20.1) [7]. Parameters for total RNA dataset: --detect_adapter_for_pe --length_required 20 --cut_right --cut_mean_quality 28; parameters for small RNA dataset: --adapter_fasta --cut_right --length_required 15 --cut_mean_quality 28; Parameters for adapter file see suppl. data sequencing_adaptors.fasta), then quality was controlled again using fastQC (v0.11.9).Mapping and counting of reads were performed differently for the total and the small RNA dataset. For the total RNA dataset, the reads were mapped with HISAT2 (v2.2.1) [18] using default parameters to the R. Norvegicus reference genome mRatBN7.2 (release 108, retrieved on 11/09/2022 from Ensembl4 ).Data was converted into sorted bam files using samtools (v1.12) [20] and reads per gene were counted using featureCounts (v2.0.1) [21] using the reference annotation mRatBN7.2 (release 108, retrieved on 11/09/2022 from Ensembl 5 ) including multimapping reads (featureCounts -M -p -T 20 -t 'exon'). For the small RNA dataset, the reads were mapped with HISAT2 (v2.2.1) [18] using default parameters to the R. norvegicus reference genome mRatBN7.2 (release 108, assembly accession GCF 015227675.2, retrieved on 06/01/2023 from NCBI 6 ). Data was converted into sorted bam files using samtools (v1.12) [20]. For counting, a miRNA specific annotation file was generated, based on the NCBI assembly GCF 015227675.2 for mRatBN7.2 (retrieved on 06/01/2023 from NCBI 7 ). The created miRNA annotation file contains a 5' and 3' miRNA entry for each miRNA, either taken from the original annotation or generated by splitting the pre-miRNA entry for the miRNA into two halves (see suppl. data rno miRNAs.gtfs). The reads per miRNA were counted using featureCounts (v2.0.1) [21] including multimapping reads (featureCounts -M -p -T 20 -t 'exon').Sequencing statistics (see Tab. 2 and Tab. 3) were extracted from fastp report files (amount of reads, read length), from featureCounts summary file. We used the percentage of reads assigned to annotated features and percentage of reads, which could not be (unambiguously) assigned to features, or not mapped to the reference at all. The percentage of risobosomal RNA (rRNA) was calculated using sortMeRNA (v4.3.6) [19] with the default sortMeRNA reference database v4.3.4 (smr v4.3 default db.fasta).The sequencing of the 32 total RNA samples (see Tab. The read duplication rate ranges from 12.88 % in sample 2d-RML-02 to 14.31 % in sample 5d-RML-03, with an average read duplication rate of 13.64 %. After trimming low quality bases and removing adapter sequences, the number of reads and the average read length decreased to about 23 million reads per sample and average read length of 184 and 174 nt for R1 and R2 reads, respectively. For all samples, about 60 % of the reads could be assigned to annotated features by using featureCounts and including multimapping reads (min. 52 %, max. 62 %). On average, 36 % of the reads could be mapped unambiguously onto the genome into regions with no annotated features, while about 4 % of the reads could not be mapped unambiguously onto the rat genome (either reads not mapping to the reference genome at all or reads mapping to multiple features). Up to 58 % of the unprocessed reads could be aligned to ribosomal RNA (rRNA), based on the sortMeRNA default database with an average of 45 % aligning to rRNA over all total RNA samples.The sequencing of the 20 small RNA samples (see Tab. removing adapter sequences, the number of reads and the average read length decreased decreased to about 82 million reads per sample and an expected average read length of 23 nt for both R1 and R2 reads.For all samples about 62 % of the reads could be assigned to annotated features by using featureCounts and including multimapping reads (between 58 -68 %). On average, 15 % of the reads could be mapped unambiguously onto the genome into regions with no annotated features, while about 23 % of the reads could not be mapped unambiguously onto the rat genome (either reads not mapping to the reference genome at all or reads mapping to multiple features). Up to 2.69 % of the unprocessed reads could be aligned to rRNA based on the sortMeRNA default database with an average of 1.33 % aligning to rRNA over all total RNA samples. The dataset from our study offers a valuable resource for understanding the quantitative relationship between hepatic perfusion and function in the transcriptomes of healthy and venous-ligated livers. It enables comprehensive analyses not only on mRNAs but also on lncRNAs and miRNAs. Given the limited knowledge about lncRNAs and the lack of general computer programs for their identification, including essential secondary structure information and protein interactions, this dataset is particularly significant.It serves as a potential source for marker genes related to regeneration capacity and potential therapeutic targets. More specifically, the reader may systematically investigate the following: