Graft-versus-host disease (GvHD) is a major cause of morbidity and mortality in hematopoietic stem cell transplantation (HSCT). Despite recent advances, the incidence remains high (20–50%) and GvHD therapy exacerbates infectious complications. Acute GvHD (aGvHD) classically develops in the first 3 months post-HSCT and can almost double the cost of a transplant. Thus, a novel, non-invasive, diagnostic test that predicts the incidence and severity of aGvHD would enable more timely prophylactic therapy, reducing morbidity, and health care costs.

MicroRNAs (miRNAs) have been demonstrated as informative biomarkers for various diseases including inflammatory conditions and cancer. miRNAs are small, non-coding single-stranded RNAs, and regulate approximately 50% of all genes, by binding to the 3′UTR and repressing translation (1). More recently, miRNAs have been identified in bodily fluids where they are protected from RNase-mediated degradation by encapsulation into extracellular vesicles (EVs) or through binding to protective proteins [reviewed in Cortez et al. (2)]. These miRNAs demonstrate a novel capacity to regulate the cellular differentiation of blood cells and immune function (3). Furthermore, distinct miRNA expression patterns are associated with specific pathophysiological conditions (4).

With regard to GvHD, miRNAs represent ideal candidates for biomarker identification as they can be assessed using accurate and sensitive technology [e.g., quantitative real-time PCR (qRT-PCR)], quantified in bodily fluids that require minimally invasive sample collection (e.g., serum/urine), and further investigated for biological function (e.g., target protein identification) that may expand upon our understanding of GvHD biology. Although the field of GvHD-related miRNA and EV research is in its infancy, recent studies have demonstrated an emerging role for miRNAs as GvHD biomarkers. Indeed, miR-155, an immune cell-related miRNA, is up-regulated in T-cells from mice developing aGvHD following HSCT, and miR-155 inhibitors resulted in decreased aGvHD severity and prolonged survival (5). Furthermore, recent murine models have shown that T regulatory cells (Tregs), critical cells in the prevention of auto-immunity, can modulate effector T-cell responses via the release of EV encapsulated miRNAs (6).

Xiao et al. previously reported that a four miRNA plasma signature (miR-423, miR-199, miR-93*, and miR-377) could predict the probability of aGvHD and was also elevated at a median of 16 days prior to disease onset (7). In addition, the miRNA signature was associated with aGvHD severity and survival (7). However, the data have not been replicated in an independent allogeneic HSCT (allo-HSCT) cohort. Furthermore, serum samples have been shown to demonstrate higher miRNA expression than plasma (8), while expression correlates well between the two fractions (9), indicating serum may be an alternative non-invasive fluid for biomarker assessment.

This study sought to replicate the prognostic and diagnostic potential of miR-423, miR-199, miR-93*, and miR-377, by assessing expression in serum samples from HSCT patients taken prior to disease onset [day 14 (D14), prognostic cohort], or at disease presentation (date of diagnosis, diagnostic cohort). To further explore the expression patterns of these miRNAs during early HSCT, expression was assessed in serum samples taken at sequential time points from pre-transplant to day 28 post-HSCT. Finally, a novel investigation of miRNA expression in the EV fraction of serum was performed at sequential, prognostic, and diagnostic time points, in order to elucidate their potential role in circulating microvesicles during aGvHD biology.

Recently, circulating miRNAs within the serum or plasma have been shown to act as robust biomarkers for a variety of diseases. In this study, expression of miR-423, miR-199, miR-93*, and miR-377 was assessed in circulating serum prior to aGvHD, at disease onset and at sequential HSCT time points, and associated with predicting disease incidence and severity. The study is initially replicatory in nature, in order to validate the results of previous reports demonstrating the potential of circulatory miRNAs as aGvHD biomarkers (7), and consequently extends these data into the novel field of EVs.

Unfortunately, GvHD has remained a frequent complication of HSCT. Diagnosis is typically based on clinical parameters confirmed by biopsy and thus, biomarkers are needed to aid in predicting disease onset and for objective diagnosis. Plasma biomarkers, particularly proteins, have shown promise (20). However, extensive validation is required to improve risk stratification or begin customized treatment, and protein-based methodology can be insensitive and unsuitable for clinical translation. In contrast, miRNAs are low in complexity, expressed specifically in different tissues, and biological stages, robust and can be accurately detected by common laboratory techniques (8, 21).

Xiao et al. reported that high expression of a four miRNA model (miR-377, miR-423, miR-199, and miR-93*) was associated with aGvHD onset and poor overall survival (7). As with all potential biomarkers, however, replication is required in independent cohorts to confirm reproducibility, especially when clinical protocols can be varied. We initially assessed the candidate miRNAs in an prognostic cohort of D14 post-HSCT serum samples and confirmed the results of Xiao et al., demonstrating significantly elevated expression of miR-423, miR-199, miR-93*, and miR-377 in patients who subsequently developed aGvHD, and their prognostic ability by ROC analysis. Xiao et al. also assessed miRNA expression in matched patient samples taken before the onset of aGvHD (2 weeks post-HSCT) and after the onset of aGvHD symptoms (6 weeks post-HSCT) and observed higher expression after aGvHD had occurred (7). This may appear to be contradictory with results of the sequential time point expression analysis performed in our study, where we assessed miRNA expression at 7-day intervals until 28 days post-HSCT and observed highest expression levels at day 7, which then decreased in all patients regardless of aGvHD status. However, it is difficult to make a direct comparison between the data sets without having a direct comparison of time points. Indeed, expression at the earlier 7-day point was not compared in the Xiao et al. study, and additionally, the last time point assessed in our sequential analysis was D28 post-HSCT, while patients developed aGvHD at a mean onset of 42 days (prognostic cohort) and 34 days (diagnostic cohort) post-HSCT. According to our data, although miRNA expression decreased in all patients between 7 and 14 days post-HSCT, levels between 14 and 28 days post-HSCT remained similar, and increased in the case of miR-377. Is it not inconceivable that if a comparison was made between miRNA expression levels pre- and post-aGvHD time points in the present study, differences may be observed. Xiao et al. also reported an association between the combined miRNA model and poor OS (7). Although a significant association with OS was not identified in the present study, the model was correlated with NRM, suggesting the miRNAs may predict mortality from causes other than relapse, such as incidence of aGvHD. In an independent diagnostic cohort comprising serum samples taken at aGvHD onset, miR-423, miR-199, and miR-93* were also elevated, further demonstrating their dysregulated expression in aGvHD and important biomarker potential. Although data demonstrated by Xiao et al. was generated using plasma samples, studies comparing the plasma, and serum fractions of blood have suggested a high degree of correlation in miRNA expression (8, 9, 22). Thus, verifying the results of Xiao et al. in an alternative sample source further highlights the reproducibility of the data.

Xiao et al. hypothesized that the candidate miRNAs may play a role in the process of donor T-cells alloreactivity against recipient tissues, or in response to damage in target cells during aGvHD, via selective inclusion into multivesicular bodies or exosomes which are released into the blood stream. In the present study, the EV fraction of serum was investigated in sequential early HSCT time points and results replicated in a larger cohort of D14 post-HSCT samples. Although disadvantages have been reported with the choice of serum compared to plasma as an EV source, such as clot formation during the serum processing procedure affecting EV release, either by trapping EVs within the clot (23) or stimulating releasing platelet-derived EVs (24), serum samples are frequently utilized for optimization of EV isolation and analysis techniques (25–27). It has been suggested that plasma can be also be a challenging material, due to high levels of clotting factors interfering with EV precipitation reagents (PRs), leading to contaminating proteins, and microvesicles. While some reports suggest no difference in EV isolation from serum or plasma (23), additional studies are required in order to fully understand how EV expression might be affected by sample processing protocols (28). In this regard, the present study utilized a commercial PR for serum EV isolation. These reagents are relatively new to the market, and their specificity for EVs remains to be thoroughly validated. In a previous study, we demonstrated the LT PR to recover smaller vesicles than a competitor product, indicating improved selectively for small EVs (12). Other studies have suggested that EV pellets isolated using PR are free from contaminating cell organelles, including apoptotic bodies (29) and that the method recovers more EVs than ultracentrifugation (UC), chromatography, or magnetic beads, with higher EV RNA and protein purity (30). Interestingly, a report specific to serum samples demonstrated low levels of protein-bound miRNA contamination in EVs isolated using PRs (27). Although UC has historically been regarded as the gold standard for EV isolation, the practical application of this method is limiting with regard to the sample volumes required, length of isolation, and requirement for specialist equipment. To this end, PRs have been evaluated as a practically applicable alternative to UC in a clinical setting (31). However, despite the acceptance of PRs for EV isolation, caution must be applied and it would be incorrect to claim that the isolated EV pellets are entirely pure in nature.

The EV aspect of this study is novel and extends upon the hypotheses proposed by Xiao et al. EVs are known to regulate cellular differentiation of blood cells, metabolic pathways, and modulate immune function (3) and play a role in antigen presentation either directly, or indirectly via transfer to antigenic peptides or by cross-dressing APCs (32). Upon activation, T lymphocytes produce large amounts of EVs that bear TCR from the pool of activated complexes (33) and it is possible that sub-sets of T-cells may be further defined by their EV miRNome. Indeed, 20 intracellular miRNAs were shown to distinguish between sub-populations of CD4⁺ T-cells, defining the development, and differentiation of this lineage (34). Furthermore, CD4⁺CD25⁺Foxp3⁺ Tregs released EVs following activation in a mouse model (35) and CD8⁺CD25⁺Foxp3⁺ T-cells secreted EVs that inhibited DC-induced CD8⁺ CTL responses (36), demonstrating immune suppressive capacities for these EVs.

In the present study, expression of the candidate miRNAs (miR-423, miR-199, and miR-93*) was lower in the EV fraction of patients who subsequently developed aGvHD. In comparison with serum, miR-93* in the EV fraction did not show a strong association with aGvHD, suggesting this miRNA may not be differentially packaged into serum EVs in an HSCT setting. Patients who did not develop aGvHD had significantly higher levels of miRNAs within the EV fraction, and we hypothesize that these miRNAs are being specifically transported in the circulation to distal target organs, where they function in an aGvHD protective capacity. However, the relationship between miRNA expression in whole serum and the EV compartment is not completely understood, and the main source of circulatory miRNA carrier in serum is still debatable. While some groups have reported that the majority of miRNAs are included within EVs (37), there is also evidence to suggest that miRNAs are more commonly co-fractionated with protein complexes, such as Argo-2 (38, 39). To add further complication, it has also been shown that the mode of transport can be specific to individual miRNAs, for example an HDL, EV, or Ago2 carrier (39, 40), and the factors determining which mechanisms are used are also unclear. Finally, although it is well recognized that specific miRNA populations are selectively sorted in to EVs (41), current research suggests four potential mechanisms for this specific packaging [nSMase2-dependent pathway, hnRNPs-dependent pathway, the 3′end of the miRNA sequence-dependent pathway, and the miRISC-related pathway (42)], and understanding of these processes in a cell-, tissue-, bio-fluid-, and disease-specific context is unexplored. Altogether, while we have hypothesized explanations for the disparate miRNA expression between serum and EVs, this remains an area of ambiguity, but likely reflects the underlying function of individual miRNAs.

However, identifying the biological function of miRNAs can be challenging, due to their short seed sequence potentially targeting hundreds of mRNA transcripts. miR-423, miR-199, and miR-93* have suggested roles in inflammation, tissue damage, regulation of cell proliferation, and tissue repair, but their specific functions in an aGvHD setting are yet to be elucidated (7). miR-423 has been implicated as a circulating marker for heart failure (43) and promotes cell growth by regulating G₁/S transition in hepatocellular carcinoma via p21Cip1/Waf1 (44). It has been postulated that cell cycle progression from the G₁/S phase may be a critical step for activation of immune-related cells in the pathogenesis of GvHD. Thus, miR-423 expression observed in this study may function to regulate Waf1, affecting cell cycle progression. miR-199 has also been shown to regulate cell proliferation and survival, by targeting caveolin-2, a plasma membrane scaffolding protein involved in cellular growth control, and apoptosis (45). Interestingly, caveolin-1 has been shown to regulate TCR signal strength and Treg differentiation, whereby cav-1⁻/⁻ donor T-cells cause less severe GvHD and yield higher numbers of Tregs (46). Caveolin-1 and -2 share structural similarities and are commonly co-expressed (45), however, miR-199 regulation of caveolin-1 is less well established in non-cell line material. miR-199 has also been shown to regulate mTOR in malignant cells (47–50). In GvHD, suppression of mTOR by rapamycin inhibits the generation of Th17 cells and enhances TGF-β-induced generation of FoxP3⁺ Tregs (51). A low Th17/Treg ratio correlates with severe clinical and pathological GvHD, apoptosis intensity, and TNF-α expression (52). Accordingly, high levels of Th17 cells are associated with lethal GvHD in mice, hallmarked by extensive pathologic cutaneous, and pulmonary lesions (53), while a high percentage of Treg cells is known to ameliorate aGvHD (54). Taken together, this may suggest a protective role for miR-199, whereby increased EV expression leads to target cell mTOR suppression and a downstream reduction in aGvHD severity. miR-93* has been demonstrated to promote tumor growth and angiogenesis via targeting of integrin-B8 (encoded by ITGB8) (55). Mouse CD4⁺Foxp3⁺ Tregs express integrin-B8, which acts as a marker for thymic-derived Tregs. Functionally, Tregs from ITGB8 knockouts (KO) exhibit normal suppressor function in in vitro and in vivo models of colitis, but fail to provide TGF-β1. ITGB8 KO Tregs express higher surface TGF-β1 levels, consistent with defective processing (56), and ITGB8 KO mice demonstrate no suppressive ability against the innate or adaptive immune response and fail to reduce the level of inflammatory cytokines (57). In breast cancer cell lines, Liu et al. demonstrated that miR-93* induces mesenchymal epithelial transition, associates with downregulation of multiple stem cell regulatory genes including JAK1, STAT3, and EZH1 (58). Activated JAK1 is required for T-effector cell responses in inflammatory disease, and inhibition leads to improved survival of mice following aGvHD, reduced histopathological grading, and lower serum pro-inflammatory cytokine levels (59). Associations with STAT3 may also have implications in aGvHD pathology, whereby the STAT3 transcription factor promotes Treg instability and is increased prior to aGvHD onset (60, 61). EZH1 is also implicated in mediating GvHD, whereby conditional loss of EZH2 in donor T-cells inhibits GvHD in mice, specifically impairing T-cell differentiation into IFN-gamma producing effector cells (62). Studies in kidney have shown that miR-93* also targets vascular epithelial growth factor-A (VEGF) (63), a cytokine associated with hematopoietic cancer outcome, immune modulation, and tissue repair (64, 65). With regard to GvHD, VEGFA levels have been associated with incidence of aGvHD (66), relapse, and non-relapse mortality following HSCT (64). miR-377 has been shown to suppresses cell proliferation by targeting CDK6, a cyclin-dependent kinase that regulates G1/S progression, and is highly overexpressed in patients with GvHD (67). Interestingly, it was demonstrated that miR-377 expression is regulated by CyA, an immunosuppressant drug commonly administered for GvHD prophylaxis, and that miR-377 inhibition resulted in reduced CyA-mediated apoptosis (68). It would be interesting to assess the effect of CyA on miR-377 expression in serum. In the present study, miR-377 was not replicated as a reproducible predictive biomarker and interestingly, all Newcastle patients with available information on aGvHD prophylaxis received CyA.

The increased miRNA expression observed at day 7 post-HSCT was an interesting finding of this study, particularly as levels remained higher in the patients who subsequently developed aGvHD. We did not detect any significant differences in the conditioning regimens between patients who developed aGvHD compared with those who did not, and so this cannot explain the data. We hypothesize that elevated miRNA expression at day 7 in both aGvHD and no aGvHD patients is a reflection of an elevated inflammatory reaction, in response to the initiation of HSCT conditioning regimens, and the HSCT procedure. It is well established that mucosal barrier injury, induced by intensive cytotoxic therapy prior to HSCT, initiates the release of pro-inflammatory cytokines, chemokines, antimicrobial peptides, and the acute phase protein, LPS-binding protein (LBP) (69, 70). Interestingly, serum expression levels of inflammatory markers (CRP, IL-8, and LBP) peak 20 days following cytotoxic conditioning therapy, which corresponds to 1-week post-HSCT (70, 71). Furthermore, a range of miRNAs are significantly elevated in serum post-chemotherapy (72). miRNAs have been established as key regulatory factors in wide ranging elements of immune function including homeostasis, response, tolerance, and auto-immunity. Furthermore, their functionality may be achieved by multiple arrangements, including a linear model (one miRNA controlling one target), divergent model (single miRNA controlling multiple targets with distinct biological outcomes), network model (coordinated regulation of multiple targets with one outcome), and convergent model (many targets coordinated by cognate miRNAs in a linear manner, contributing additively to a specific response) (73). Taken together, this highlights the significance of miRNAs as regulatory elements in the context of immune response, and the complexity surrounding trying to decipher their expression and specific function in relation to immune response activity.

As discussed, the miRNAs examined in the present study have previously been associated with inflammation, tissue damage, regulation of cell proliferation, and tissue repair and thus, it is not inconceivable that they may be among the spectrum of miRNAs elevated post-conditioning. Although it is not clear why these miRNAs remain elevated in the sub-group of patients who develop aGvHD compared with those who remain aGvHD-free, it is possible that this is a reflection of the escalating inflammatory cascade that occurs in patients who develop aGvHD. Indeed, elevated circulatory miRNA expression may initially be driven by the conditioning and transplant procedure-initiated inflammatory reaction, which subsequently manifests into a cytokine storm and tissue damage, induced by alloreactive T-cells of donor origin in patients who develop aGvHD, leading to sustained miRNA expression. Conversely, the inflammatory reaction diminishes in patients who do not develop aGvHD and miRNA expression may also be reduced, if not further stimulated by the allogeneic response mediated by donor T-cells. This would be an interesting area for further investigation. Although Sun et al. performed studies in mice to examine miRNA expression patterns in allogeneic T-cell responses, they did not detect significant expression of the miRNAs reported in the present study, despite homologs being present on the miRNA array (74). However, their study focused on allostimulated T-cells in vitro, while the present study concentrated on circulatory miRNA expression in vivo. Furthermore, differences in miRNA expression and function between mouse and clinical models cannot be ruled out. However, this does highlight the importance of comprehensive miRNA expression profiling in the aGvHD setting, which takes into account both the full array of recognized miRNAs and the heterogeneity in patient transplant protocols. Indeed, in a recent study we performed global expression profiling of n = 799 mature miRNAs in the serum of aGvHD patients (n = 12) at aGvHD diagnosis (75), which incorporated an expanded repertoire of miRNAs compared with the n = 345 species profiled by Xiao et al. (7). Although differential expression of the miR-423 and miR-199 (miR-93* was not present on the codeset) candidate miRNAs was confirmed, additional species with highly significant differential expression were also identified, suggesting that a panel of signature miRNA biomarkers may be a more potent predictor of aGvHD incidence and severity for this complex disease.

In conclusion, this study has replicated circulating miR-423, miR-199, and miR-93* as diagnostic and prognostic biomarkers for aGvHD in independent cohorts. Principal component analysis of miR-423, miR-199, and miR-93* confirmed the utility of the miRNA triad in predicting occurrence of aGvHD. In addition, for the first time we have demonstrated the expression of these miRNAs in the EV fraction, and results suggest patients who do not develop aGvHD have a higher level within serum EVs at D14, proposing a novel predictive biomarker role for EVs. The source of dysregulated miRNA expression remains unclear, particularly between donor-derived hematopoietic stem cells and patient-derived immune response cells, and this represents an important area for further investigation. Further assessment of the biological role of these miRNAs, particularly in the EV compartment, should inform on their contribution to the development of aGvHD.

This study was carried out in accordance with the recommendations of the Newcastle and North Tyneside I Research Ethics Committee and the Ethics Committee of the Medical University of Vienna, Austria. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

RC designed and performed the experiments, evaluated the data, and wrote the manuscript; JN assisted with experiments and drafting of the manuscript; MJ performed replication experiments in diagnostic samples and LB helped establish qRT-PCR protocols; KP advised and performed statistical analysis; CL performed clinical data collection and advised on PCA analysis; MC and HG contributed to clinical samples, data collection, and preparation of the manuscript; and AD developed the overall concept, supervised the research, and prepared the manuscript. All authors approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.