MicroRNAs in blood act as biomarkers of colorectal cancer and indicate potential therapeutic targets

Association studies have linked alterations of blood‐derived microRNAs (miRNAs) with colorectal cancer (CRC). Here, we performed a microarray‐based comparison of the profiles of 2549 miRNAs in 80 blood samples from healthy donors and patients with colorectal adenomas, colorectal diverticulitis and CRC at different stages. Confirmation by quantitative real‐time PCR (RT‐PCR) was complemented by validation of identified molecules in another 36 blood samples. No variations in miRNA levels were observed in samples from patients with colorectal adenomas and diverticulitis or from healthy donors. However, there were 179 CRC‐associated miRNAs of differential abundance compared to healthy controls. Only three – miR‐1225‐5p, miR‐1207‐5p and miR‐4459 – exhibited increased levels at all CRC stages. Most deregulated miRNAs (128/179, 71%) specifically predicted metastatic CRC. Pathway analysis found several cancer‐related pathways to which the miRNAs contribute in various ways. In conclusion, miRNA levels in blood vary throughout CRC progression and affect cellular functions relevant to haematogenous CRC progression and dissemination. The identified biomarker and therapeutic candidates require further confirmation of their clinical relevance.


Introduction
Colorectal cancer (CRC) is a leading cause of cancerrelated death worldwide [1]. The 5-year survival rate of nonmetastatic CRC (UICC stages I-III) ranges between 60% and 95% and falls to < 15% for patients with metastatic CRC (mCRC, UICC stage IV) [1,2]. Thus, there is a need for developing improved CRC screening and treatment options to reduce CRC mortality [2,3] MicroRNAs (miRNAs) are potent post-transcriptional gene expression regulators and appear to be involved in virtually all features of CRC, including initiation, progression and its metastatic spread [4]. Association studies have linked alterations of blood-derived miR-NAs with CRC, but they assessed only a limited number of miRNAs with poor replicability of putative biomarker candidates across studies [5,6]. Comprehensive assessments of miRNA abundance changes in blood from CRC patients are lacking. There are only studies reporting about analyses on some miRNA molecules each [5][6][7][8][9]. Also, the blood-based miRNAs reported to be deregulated in CRC were not controlled for confounding effects by premalignant and/or inflammatory colorectal conditions, such as colorectal adenoma (CRA) or diverticulitis (CRD) [5,9]. Moreover, miRNA profiles may diversify during CRC progression and reflect in blood the CRC progression level. No study has yet provided an assessment of miRNA alterations in blood at different CRC stages.
The human genome encodes over 2500 miRNAs, which function in complex regulatory networks, often in reciprocal interactions with the targets and pathways that they regulate [10]. Current technologies provide platforms to address this complexity as they enable detection of all known human miRNAs in a single assay [11]. While biomarker discovery focused on miRNAs with consistently altered abundance across CRC stages, distinct miRNA levels during progressive stages may reflect how CRC dissemination is represented in blood. Apart from the easier access to blood rather than tissue biopsies, blood samples may provide information regarding compartment-related functions of miRNA alterations in haematogenous CRC dissemination. To gain an insight into the biological meaning of miRNA dysregulation in blood, bioinformatics tools that interrogate knowledge databases can help identifying affected targets and pathways at the level of large miRNA datasets [12]. Together, these methodologies may reveal previously unknown prometastatic effects of miRNA dysregulation in haematogenous spread and identify tumour vulnerabilities for the design of novel antimetastatic strategies.
In the present study, we used a human miRNA analysis microarray to compare blood profiles of 2549 miRNAs in healthy control donors and patients with CRD and CRA as well as CRC at different UICC stages (I/II, III and IV) with the aim of assessing candidate biomarker miRNAs for blood-based diagnostic tests common to all CRC stages. We also tested for stage-independent but CRC-specific alterations. Further, we subjected to bioinformatics analysis the miR-NAs that exhibited changed abundance levels in the blood for identifying potential target molecules and pathways that may be affected by particularly mCRCspecific alterations in the miRNA profile.

Ethics statement
Research was performed in accordance with the Declaration of Helsinki of the World Medical Association. The study was approved by the ethics committee of the € Arztekammer Hamburg (positive ethics vote PV4594 of 28 November 2013). All patients and healthy donors gave written informed consent to use their samples for research prior to their inclusion in the study.

Sample collection
The participants for this prospective exploratory single-centre study were recruited at the Asklepios Hospital Barmbek in Hamburg, Germany, between January 2014 and December 2016 as well as between November 2020 and February 2021. The characteristics of the participants and the criteria for diagnosis and cohort inclusion are presented in Table 1. From each study participant, 5 mL peripheral blood was drawn into PAXgene Blood RNA tubes (BD, Franklin Lakes, USA). Blood samples were snap-frozen in liquid nitrogen directly after sampling and subsequently stored at À80°C until used for RNA isolation.

RNA isolation and miRNA microarray analysis
RNA isolation was performed as described in detail earlier [13][14][15]. In short, 5 mL blood was centrifuged at 5000 g for 10 min at room temperature. Total RNA including miRNA from the cell pellet was isolated using the miRNeasy kit (Qiagen, Hilden, Germany) according to the manufacturer's instruction. RNA integrity was evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Only samples with an RNA integrity number of at least seven were used for further analyses. The total RNA prepared from individual samples was analysed at the Genomics Core Facility of the German Cancer Research Center (DKFZ) in Heidelberg on the Agilent Human miRNA Microarray Release 21.0 according to the detailed manufacturer's protocol that is available from Agilent and includes process controls. Briefly, fluorescently labelled miRNA was prepared according to the Agilent miRNA Complete Labelling and Hyb Kit protocol. Hybridisation was performed in a Sure-Hyb chamber (Agilent Technologies) at 55°C for 20 h. Microarrays were scanned using the Agilent Scanner G2505C. After quantile normalisation, fold changes were calculated based on mean values of technical replicates. In total, 2549 human miRNAs were analysed for differences in their blood levels.

Statistical analysis
Statistical analyses were performed using the Chipster microarray data analysis software (http://chipster.csc.f i). After RMA normalisation and log 2 -transformation of expression values, fold changes were calculated based on the mean values of the replicates. Single miRNA analyses for expression variations were carried out using t-tests (unpaired, two-tailed) for pairwise comparisons after verifying approximate normal distribution using empirical Bayes method. The resulting pvalues were adjusted for multiple testing using Benjamini-Hochberg's adjustment. A log 2 -fold change (FC) > 0.2 and an adjusted P-value < 0.05 were considered as cut-off criteria for significant up-or downregulation of differentially abundant miRNAs.

Evaluation by quantitative real-time PCR
Each RNA sample was checked for genomic DNA contaminations that could obscure the measurement. One microgram of total RNA was converted to cDNA using the miScript Reverse Transcriptase mix (Qiagen) following the manufacturer's instructions. Quantitative RT-PCR was performed in triplicate with a LightCycler 480 instrument (Roche Diagnostics, Mannheim, Germany) using the miScript SYBR Green PCR Kit according to the manufacturer's protocol. Molecules of the QuantiTect Primer Assays (Qiagen) were used as primers: hsa-miR-1225-3p, hsa-miR-1260_1, hsa-miR-4516_1 and hsa-RNU6-2_11, the last used for normalisation purposes as suggested by Qiagen. Data were analysed using the LIGHTCYCLER software (Roche Diagnostics).

Ingenuity pathway analysis
Bioinformatics analyses for all up-and downregulated miRNAs in the microarray dataset of mCRC patients in comparison with healthy controls were performed using the Ingenuity Pathway Analysis (IPA) software (Qiagen). Microarray-based miRNA expression data were sorted by the log 2 -fold change, uploaded into IPA for core analysis and then overlaid with the largescale molecular network implemented and available in the Ingenuity Knowledge Base. Downstream effect analyses were carried out to identify those cellular functions, targets, pathways and networks that are likely to be affected by the patterns of differentially abundant miRNAs. These analyses provided predictions about the activation state of each function, target, pathway and/or network, calculated as activation Z-scores ([Z] > 2 is considered significant) [12].

Study population
The initial study participants included healthy subjects as control group (cohort 1) and newly diagnosed patients with the premalignant and inflammatory colorectal conditions diverticulitis (CRD; cohort 2) or colorectal adenoma (CRA; cohort 3), as well as CRC at stages UICC I/II (cohort 4), UICC III (cohort 5) or UICC IV (cohort 6). They were recruited for this exploratory study between January 2014 and December 2016. The baseline characteristics of all participants and the criteria for diagnosis and cohort inclusion are presented in Table 1. Notably, the absence of CRC was confirmed by colonoscopy in all CRC-free individuals (cohorts 1-3), and the presence of CRC was proven histologically in all CRC patients (cohorts [4][5][6]. No treatment had been administered to any of the blood donors prior to sampling, specifically no antibiotic therapy in cohort 2 (CRD), no polypectomy in cohort 3 (CRA), no resection in cohorts 4 and 5 (CRC, UICC I/II and III) and no systemic therapy in cohort 6 (mCRC, UICC IV). From each study participant, peripheral blood was collected, snap-frozen in liquid nitrogen directly after sampling and stored at À80°C until used for RNA isolation. In the analysis, 80 samples met the strict quality requirements of the profiling analysis, including samples obtained from patients with CRA (n = 8), CRD (n = 8) and CRC at different stages (I/II, III and IV; for each group, n = 16) and from 16 healthy donors (Table 1).

miRNA level profiles
Sample analysis followed a robust process that has been used in various analyses [13][14][15], including a study that looked at more than 1000 blood samples from patients with different cancers and noncancer diseases. Surprisingly, no significant variations in miRNA levels compared to the healthy controls (cohort 1) were detected in samples obtained from patients with colorectal adenoma and diverticulitis (cohorts 2, 3). This indicates that the identified CRC-associated miRNA alterations were not confounded by premalignant or inflammatory colorectal conditions. By contrast, a total of 179 (7%) miRNAs with significantly changing levels (92 up and 87 down) were obtained from colorectal cancer patients (cohorts 4-6). There were actually rather large differences between UICC stages I/II, III and IV regarding the number and pattern of differentially abundant miR-NAs in comparison with samples from healthy blood donors (Tables S1-S3, respectively). Most CRCassociated deregulated miRNAs (71.5%, 128 of 179, 63 up and 65 down) were found at stage IV. The degree of specificity to UICC stages of the significantly varying miRNAs is documented by the small number of miR-NAs that exhibited change at more than one stage (27 of 179; 15%; Fig. 1). Merely three miRNAs (1.7%, 3 of 179) -miR-1225-5p, miR-1207-5p and miR-4459were detected to exhibit increased levels across all UICC stages. As a control, miR-1225-5p and two randomly selected deregulated miRNAs were analysed by RT-PCR in all samples from healthy donors and tumour patients for confirmation of results obtained in the microarray analyses ( Fig. 2A). To prove that the results hold true also in an independent set of samples, we performed RT-PCR on miRNA isolated from blood of a second group of participants, who were recruited between November 2020 and February 2021. miRNA was isolated from blood of 8, 8 and 10 patients with stage I/II, stage III and stage IV CRC, respectively, as well as miRNA from 10 healthy blood donors (Table 1). Repeating the RT-PCR, we obtained similar results, although a better discrimination could be observed (Fig. 2B).

Pathway analysis
Ingenuity Pathway Analysis (IPA) was applied to the dataset of the stages I/II, III and IV in order to define pathways and molecules that might be affected by the observed changes in miRNA levels. Although less informative, we also performed a target scan for all miRNAs that exhibited variations in their abundance (Table S4). While stage I/II samples yielded several pathways that exhibit an association with cancer (Fig S1), only a single such network was identified analysing the stage III data (Fig S2). This result is not surprising, since far fewer miRNAs were found to be specific for stage III.
Of particular interest was stage IV (mCRC). According to the number of specifically varying miRNAs, several pathways associated with cancer were identified, and they were also the most strongly affected ones ( Table 2). Using existing functional annotations of the genes targeted by the miRNAs, the functions and diseases were defined, in which each network plays a role.
A score was computed that indicates the relevance: a score of 2 is considered significant. Actually, the top scores obtained with the miRNA networks shown range from 22 up to 33 (Table 2), highlighting the importance to mCRC of the miRNA variations. Networks involving SMAD2/AKT (Fig. 3A) and CDKN1A (Fig. 3B) belong to ones that are affected by the observed miRNA patterns. The roles of SMAD2/AKT and CDKN1A in the progression of CRC are well established [16][17][18][19][20][21]. Finding them as a result of the miRNA profiling supports the high relevance of the observed miRNA variations with respect to CRC. In the other two most prominent networks, the link to cancer is more indirect for the respective genes although the score of their relevance to CRC is basically identical to that of SMAD2/AKT and CDKN1A ( Table 2). The genes encoding for Neuregulin 1 (NRG1, Fig. 3C) and chemokine C-X-C ligand 12 (CXCL12, Fig. 3D) were identified as being strongly affected by the detected miRNA increases and decreases. Both genes are located central in their respective regulative network. NRG1 is one of four ligands (NRG 1-4) for the human epidermal growth factor receptor family (HER 1-4), which in turn is strongly implicated in the growth and spreading of cancer cells [19,22]. CXCL12 is the main ligand of chemokine receptor 4 (CXCR4) and overexpressed in metastasis of various cancers including CRC [23,24]. Based on the identification of genes/proteins that correspond to the mCRC-specific miRNA patterns (Table 2), also relevant signalling pathways could be determined.   Table 2. List of the most strongly affected regulative networks in metastatic colorectal cancer (mCRC). A pathway analysis was performed based on the observed miRNA variations. Ranking was according to the number of miRNAs found to be involved in a network. Functional and disease annotations are given next to a score that indicates relevance; a score > 2 is considered significant. The molecules involved in the networks are listed. In bold, the regulated miRNAs are shown; the direction of changeup or downis indicated by arrows.

Discussion
In this study, we compared miRNA abundance profiles in blood samples taken from patients and a cohort of healthy individuals. The study did not investigate plasma molecules but analysed the entire miRNA content of a blood sample, including that of peripheral blood mononuclear cells (PBMCs) or red blood cells. It has been suggested that the latter even act as miRNA repositories [25]. A large number of studies have documented the feasibility of this approach [e.g. [13][14][15]. As no analysis has yet been performed that compares plasma from blood profiles, it is difficult to evaluate to what extend these two types of miRNA profiles may be distinct or overlapping. All cells are receiving and releasing miRNAs and are therefore likely to both be affected by and contribute to the miRNA composition in plasma [e.g. 26].
To our surprise, we did not find altered miRNA levels in blood from non-CRC patients with diverticulitis and colorectal adenomas as compared to healthy donors. The lack of any changes is remarkable, as earlier studies on various diseasescancerous and noncancerousdid consistently identify level variations in blood [9,13]. Why this is different for colorectal adenomas and diverticulitis requires further studies. However, the result indicates that the identified CRC-associated  Top diseases/functions   4  AKT1, CHID1,CXCL12, DZIP1, GPR162, GPR171, GSK3B, LPAR2,  MOGAT3, MORN5, MTMR7, PLCL1, PPP1R7, PRSS22, SEC14L2 Cancer, gastrointestinal disease, organismal injury and abnormalities miRNA alterations are not confounded by premalignant or inflammatory colorectal conditions. Another important finding was the only small overlap of deregulated miRNAs across the three CRC stages: 152 of the 179 deregulated miRNAs (84.9%) were specific for one CRC stage only; 24 (13.4%) changed in two CRC stages. Only three (1.7% of 179 deregulated and 0.1% of all 2.549 tested miRNAs) -miR-1225-5p, miR-1207-5p and miR-4459were specific for CRC at all stages. A recent meta-analysis about 617 miRNAs, which had been reported to be deregulated in 3.454 cases of CRC, did not identify a single miRNA that was common to all CRC samples [5]. However, the studies included in the meta-analysis relied on a mostly small number of preselected miRNAs. Increased miR-1207-5p levels have not yet been reported in cancer patients at all, and only one study each has mentioned increased levels of miRNA-1225-5p in breast cancer patients [27] and of miR-4459 in ovarian cancer [28]. The data (Table S4) suggest that the 179 miRNAs may have functions in CRC pathobiology when circulating in blood. Particularly the stage-specific miRNAs but also the molecules common to more than one CRC stage could have potential for blood-based diagnostic tests as a supplement to existing CRC screening processes [2].
We considered the selection of a best performing set of miRNAs for the establishment of a miRNA signature for diagnostic purposes. While there are numerous reports, in which this was done, we actually refrained from doing so. While sensitivity and specificity values could be calculated in principle and an optimal signature be established, including even a validation in the second set of 36 samples, this would nevertheless be a more academic activity with little relevance to diagnostics. The sample size is too small for producing meaningful data in this respect. Towards a potential clinic use eventually, such an analysis is called for, of course. However, substantially more samples and a multicentre approach are required to such end.

NRG1-mediated HER3 activation inhibits migration and growth of cancer cells in preclinical models
The blood circulation is crucial for CRC dissemination, which is powered by circulating tumour cells (CTCs) as potential seeds [36]. However, only a small CTC fraction of about 0.01% with a distinct phenotype can survive in the blood circulation and give rise to distant metastasis [37]. miRNAs contribute to relevant CTC programming, which could increase plasticity, motility, mobility, invasiveness and migratory traits required for intra-and extravasation [38]. miRNAs are emerging as mediators of intercellular communication to reprogram recipient cells and seem to play a key role for premetastatic niche formation at distant organ sites [39,40]. It is an interesting hypothesis that the mCRCspecific miRNAs in blood may function as paracrine mediators in cell-to-cell signalling targeting the CTC population. Uptake of miRNAs could expand the CTC fraction capable of haematogenous spread. In support of this hypothesis, all affected proteins predicted on the basis of our mCRC-specific deregulated miRNA dataset, most significantly NRG1 and CXCL12, are part of the pathways involved in cancer signalling and stem cell pluripotency.

Conclusions
This study provides leads for improved biomarker discovery for CRC screening, but it also widens the horizon for the design of novel antimetastatic therapies. Since our results have been obtained by an exploratory single-centre analysis, they warrant confirmatory testing in an independent multicentre validation. Further, clinical and functional studies will be required to determine precisely the origin and function of miRNA alterations in blood and to verify predicted causality relations. However, this is the first study providing blood expression profiles in CRC that is representative of the whole known human miRNA landscape. Moreover, it is first in providing a comprehensive assessment of miRNA alterations in blood at different CRC stages. The altered miRNA profiles were controlled for non-CRC colorectal confounders and selected by comparison to healthy controls defined by inconspicuous colonoscopy.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. IPA maps of the top four signalling pathways that correspond to the observed miRNA variations in stage I/II CRC. Fig. S2. IPA map of the signalling pathway that corresponds to the observed miRNA variations in stage III CRC. Fig. S3. Map of the signalling pathways that correspond to the observed miRNA variations in stage IV CRC (mCRC). Table S1. Blood-derived miRNAs that were down or up in abundance in colorectal cancer patients of UICC stage I/II (n = 16) compared with healthy controls (n = 16). Table S2. Blood-derived miRNAs that were down or up in abundance in colorectal cancer patients of UICC stage III (n = 16) compared with healthy controls (n = 16). Table S3. Blood-derived miRNAs that were down or up in abundance in colorectal cancer patients of UICC stage IV (n = 16) compared with healthy controls (n = 16). Table S4. Protein molecules that are probably affected by the miRNAs exhibiting varying blood levels.