The role of biomarkers in the diagnosis and risk stratification of Acute Graft vs. Host Disease (aGvHD): A systematic review

Abstract

Allogeneic hematopoietic cell transplantation (HCT) is an increasingly used curative modality for hematological malignancies and other benign conditions. Attempts to reduce the morbidity/mortality and improve survival in patients undergoing HCT are crucial. The ability to diagnose acute graft versus host disease (aGvHD) in a timely manner, or even predict aGvHD prior to clinical manifestations, along with the accurate stratification of these patients are critical steps to improve the treatment and outcomes of these patients. Many novel biomarkers that may help in achieving these goals have been studied recently. This overview is intended to assist clinicians and investigators by providing a comprehensive review and analytical interpretation of the current knowledge concerning GvHD and biomarkers likely to prove useful in diagnosis and risk-stratification of this condition, along with the difficulties that hamper this approach.

Introduction

Acute graft-versus-host disease (aGvHD) is a major complication of allogeneic hematopoietic-cell transplantation (allo-HCT), occurring in up to 60% of the transplant recipients, with well described risk factors including human leukocyte antigen (HLA) mismatch, age, stem cell source, gender disparity, and conditioning regimen[1]. Acute GvHD is a systemic disorder driven by donor T-cells with a pleomorphic clinical presentation involving multiple target organs, including the skin, liver and gastrointestinal (GI) tract [2]. The diagnosis of aGvHD is clinical and pathologic confirmation is helpful but lacks positive and negative predictive value[3–5]. Immunosuppressive therapy with steroids is the first line therapy for clinically significant GvHD. Mortality and morbidity from high dose systemic immunosuppression is significant and no other treatment to date added to upfront steroids has proven beneficial[6]. Although the clinical diagnosis can be readily made basis in patients with a classical presentation, some cases prove challenging[7]. Currently, the diagnosis of aGvHD can be made by combining the clinical impression, high pre-test likelihood of aGvHD, exclusion of other competing disorders, along with histological examination of the target tissue. Biopsy alone is not the gold standard for diagnosis due to the false negative (patchiness of the disease, absence of the typical changes at early stages) and false positive results (residual conditioning regimen toxicity, infections)[3]–[5]. Furthermore, other causes for a patients’ symptoms may be found in many cases, confounding the diagnosis of aGvHD [8]–[12]. Once diagnosed, the severity of aGvHD has historically been graded using the Keystone Consensus criteria[13] or CIBMTR criteria[14]. Grading is an important step to determine the appropriate management of aGvHD and to determine prognosis and response to treatment. It has been well recognized that current grading systems overgrade some patients with a high likelihood of responding to immunosuppressive therapy and can not predict who will respond to steroids with certainty[15]. There are several limitations to current prognostic grading systems: despite the general relationship to outcomes, inter-observer errors can occur due to subjective biases, initial grade may not reflect peak grade and the time to response after therapy initiation is not accounted for. Therefore, many patients who are classified as standard-risk have their treatment fail while others classified as high- risk are over-treated. A recent update to the acute GvHD grading system may help to better stratify patients’ non-relapse mortality risk but could still be plagued by inter-oberserver biases[15] with significant variability across BMT centers[16]. The use of biomarkers can add to prediction accuracy and eliminate some of these problems.

High-dose systemic steroids are the standard first-line therapy for patients with grade II or higher aGvHD[6], [17]. However, practices differ among institutions with respect to initial steroid dose, the use of additional immunosuppressive agents, and the approach to steroid tapering after initial response. In general, therapy of clinically suspected aGvHD is started before diagnosis is confirmed and potentially before peak grade. Such an approach may expose the already immunosuppressed patients to unnecessary systemic steroids with significant infectious and non-infectious complications. In fact, a study conducted to reduce the dose of initial steroid therapy for grades I – II acute GvHD observed no difference in outcome between standard dose 2mg/kg and reduced dose 1mg/kg arms[18]. Approximately half of patients have complete responses to steroid by day 28 after therapy initiation[19]. Steroid refractory aGvHD is associated with high transplant-related mortality and low overall survival, even without relapse of the underlying disease that required their transplant[20]. There are no standard second line therapies for steroid-refractory aGvHD, and responses vary, but based on small retrospective and phase I/II studies responses are around 30 – 50% with six month NRM approximately 50%[6]. The hope of a biomarker based grading system has the potential to more accurately stratify patients based on risk of failing to respond to steroids, to alternative treatment approaches, or in those that fail upfront therapy, may be used to help choose additional lines of therapy.

Although mortality related to GvHD has been reduced in recent years [21], [22], aGvHD remains a major cause of transplant-related mortality (TRM). Acute GvHD represents the primary limitation to more widespread use of allogeneic HCT as a potentially curative modality for patients with malignant and non-malignant diseases. The field of biomarker research may provide for more accurate grading/risk stratification and identification of patients at higher risk for refractoriness to therapy or GvHD progression. Furthermore, the treatment of aGvHD has recently evolved from a one-size-fits-all approach to a more refined strategy based on predicted outcomes. Patients who are predicted to have low-risk aGvHD may benefit from lower doses and shorter courses of immune suppression. In addition, since not all cases of aGvHD progress in the same way or have the same outcome, the therapy should be tailored not only to the severity of the disease but also to the predicted rate of progression. As a result, many researchers have examined whether adding novel plasma biomarkers levels at different time points before and after transplantation can add to the prediction accuracy when compared to other prognostic tools. Timely recognition of patients at high risk for aGvHD or who would likely demonstrate resistance to steroids early in the course of their transplantation may lead to a more stringent monitoring, preventive care and early introduction of alternative and more effective immunosuppressive treatments earlier in the course of treatment.

It is reasonable to assume that plasma proteins involved in the complex pathophysiology of aGvHD might be altered in these patients. For the past 20 years, various groups have been investigating potential biomarkers to enhance the early and more accurate diagnosis and risk stratification of patients with aGvHD. Recent research has applied proteomics technologies to identify aGvHD biomarkers. This has led, in a short period of time, to the identification of novel biological pathways and biomarkers predictive of and associated with aGvHD [23]. Nevertheless, no single biomarker nor a panel of biomarkers, has been validated for clinical use via large multicenter trials. In this article, we summarize the current knowledge of promising diagnostic and prognostic aGvHD biomarkers and analyze the supporting data available in the literature.

Review Design

We searched PubMed and Medline up to December 31, 2015 to identify studies evaluating biomarkers in the setting of aGvHD. Each biomarker (microRNA, ST2, TNF receptor 1, IL-7, sBAFF, REG3α, S100, TIM-3, CK-18, HGF, Elafin) was searched separately as well. Only full-text articles published in English were considered. The primary search was conducted using the terms “Graft-versus-host disease” and “biomarker” excluding “Reviews”. Relevant references in the publications identified were also reviewed. Eligible studies included clinical studies with more than 5 patients. Studies investigating the diagnostic and prognostic value of transcriptomic and proteomic biomarkers were reviewed. Biomarkers that were evaluated in at least 2 independent studies will be discussed. The primary statistical outcomes were sensitivity/specificity, positive/negative predictive values, area under the curve, and hazard ratios. The main outcomes of the remaining preclinical and clinical studies were reported in the table but not discussed in the text.

Diagnostic and Prognostic Biomarkers

The biomarkers of aGvHD that have been discovered so far can be classified in a variety of ways: target (systemic or particular organ/tissue) or via their biophysical properties. Systemic biomarkers lack organ specificity for skin- or GI-type aGvHD. These biomarkers rise in response to systemic injury rather than specific tissue damage. On the other hand, organ-specific biomarkers are expressed by target organs rather than the effector cells that are damaging all tissues. Identifying markers that are target-tissue specific has been technically challenging due to the cellular heterogeneity of tissues and the difficulty of amplifying the amount of protein required. Another way of classifying these novel biomarkers is based on their biophysical properties. Transcriptomic biomarkers are discovered by RNA expression profiling (mRNA, rRNA, tRNA and other non-coding RNA), while proteomic biomarker are discovered by the methodical studying of the protein profile of a biologic specimen. Finally, cellular biomarkers are discovered by the studying of the altered numbers and functions of several different immune cell subsets [24]–[26].

Below we review the the following systemic biomarkers (miRNA, ST2, and markers of immune activation) and organ-specific biomarkers (Reg3α, S100, TIM-3, CK-18, HGF, and Elafin) and summarize these results in Table 1.

Table 1.

A summery of the biomarkers of aGVHD classified according to their organ specificity.

		Biomarker	Studies	Type/N	Time	Assay	Conditioning regimen	GVHD prophylaxis	Outcomes
Systemic		MicroRNA (e.g. MiR-155, MiR-586)	Ranganathan et al32	Murine/Human (5)	D 21 post HCT	RT-PCR	N/A	N/A	MiR-155 is up-regulated in T-cell from aGVHD. Blocking MiR-155 may prevent aGVHD.
			Xie et al33	Human (64)	2–5w post GVHD	RT-PCR	MAC	CSA, MMF, MTX	MiR-155 is up-regulated in aGVHD following allo-PBSCT with correlation with severity.
			Wang et al34	Human (98)	D: 7, 14, 21, 30, 60, 90 & at onset of GVHD	RT-PCR	MAC	Not specified	MiR-586>2200 copies/uL at D7 may predict impending aGVHD (Sen: 87.5%. Spec: 55.0%. AUC: 0.739)
			Xiao et al35	Human (196)	16 days prior to GVHD diagnosis	qRT-PCR	MAC&RIC	Not specified	4-miRNA panel predicted the probability of aGVHD (Sen: 92%. Spec: 62%. AUC: 0.8) The panel was an independent predictor for developing aGVHD (HR: 1.478) at median of 16 d before diagnosis. The MiRNAs levels were significantly associated with severity The panel was an independent unfavorable prognostic factor for aGVHD OS (HR: 2.110) AUC of miRNA signature was higher than the AUC for sIL2Ra 2 weeks after HCT (0.86 vs 0.76)
		ST2	Vander Lugt et al43	Human (673) Bi-center	D 14 after HCT D 16, 28 after therapy of GVHD	Mass spectrometry/ELISA	MAC&RIC	CNI/MMF CNI/MTX CNI/MTX/Enbrel	High ST2 levels on D14 is associated with an increased NRM within 6 months of HCT (58% overall NRI) Patients with high ST2 levels at initiation of therapy for aGVHD were 2.3 times as likely to have resistant aGVHD and 3.7 as likely to die within 6 months compared to patients with low levels.
			Ponce et al44	Human (113)	D28 after CBT	ELISA	MAC&RIC	CNI/MMF	High D28 ST2 levels (≥ 33.9 ng/mL) were associated with increased risk of III–IV aGVHD (HR: 2.62. AUC: 0.59) and increased TRM (HR:4.2. AUC:0.75)
		Ceruloplasmin	Lv et al95	Human (98)	d-9, -1, 7, 14, 21, 28, 55–60,90–100	ELISA	MAC	CSA+MMF+MTX	Ceruloplasmin levels> 670 μg/ml on D 7, 14, 21 predicted aGVHD occurrence (Sen: 79.7%, Spec: 78.3%, AUC: 0.861). The levels failed to predict resistance to first-line therapy.
Organ-specific	Immune activation	TNFR1	Choi et al47	Human (438)	Before and at d7 after HCT	ELISA	MAC	FK506+mini-MTX	Day7 TNFR1 ratio correlated with eventual development of II–IV GVHD (HR=2.44) and TRM (HR=2.40)
			Kitko et al48	Human (82) Pediatrics	Before and at d7 after HCT	ELISA	MAC	FK506+MMF FK506+mini MTX	Day7 TNFR1 ratio correlated with severity and 1-year OS in pediatric population
			Willems et al49	Human (106)	Before and at d7 after HCT	ELISA	RIC	CNI+MMF	Day7 TNFR1 ratio correlated with grade II–IV (HR=2.2) and grade III–IV (HR=2.9) aGVHD but no impact on OS
			August et al50	Human (62) Peds & adults	Before, d0, 5, 10, 15 after HCT	ELISA	RIC&MAC	Not specified	D15 levels of TNFR1 (among other biomarkers: sCD8, sIL-2R, sCD40, sCD28) correlated with severe aGVHD (AUC=0.77, PPV=0.45, NPV=0.87)
		IL-7	Dean et al53	Human (31)	D7, 14. Month 1, 2, 3, 6, 9,12	ELISA	RIC	CSA+MTX	IL-7 level at D+14 (cutoff value of 13 pg/mL) predicted the subsequent development of aGVHD (Sens: 85.7%, Spec: 88.2%, PPV: 85.7%, NPV: 88.2%). Higher levels were strongly associated with more severe grades of aGVHD.
			Thiant et al55	Human (45)	D0,7, 14, 18, 25, 30, 60, 90	ELISA	RIC	CSA+MTX	IL-7 and IL-15 have similar kinetic, peaking on day +14 at four- to fivefold over pre-conditioning values (medians of 15.8 and 38.7 pg/ml). The occurrence grade II–IV aGVHD is associated with peak IL-7 level (HR=5.38)
			Thiant et al54	Human (40)	D0, 7,14,18,25,30,60,90	ELISA	MAC	CSA+MTX
		sBAFF	Cho et al57	Human (45)	D0, 7, 14	ELISA	MAC	CSA+MTX FK506+ MTX	sBAFF levels > 43 pg/mL at each time point had a significantly lower cumulative incidence of aGVHD (Sen: 75%, Spec: 73–82%, AUC: 0.7–0.8)
		TIRC7	Zhu et al96	Human (39)	Before and after GVHD therapy	q-PCR &ELISA	MAC&RIC	CSA+MTX	Higher levels of TIRC7 in aGVHD. Levels correlated with severity and declined markedly after therapy.
	Lower GI-specific	REG3a	Harris et al62	Human (954) Multi-center	At onset of GVHD	ELISA	High&moderate intensity	CNI/MMF CNI/MTX	REG3a (>151 ng/mL) distinguished LGI GVHD from non-GVHD diarrhea (AUC: 0.79, PPV: 95%, NPV: 34%). REG3a at onset of LGI GVHD predicts nonresponse to therapy at D28 (PPV: 51%, NPV: 76%). Higher REG3a concentrations at onset of LGI GVHD (above the median >135 ng/mL) correlated significantly with higher 1-year NRM (52% vs 33%)
			Ferrara et al63	Human (1014) Multi-center	At onset of LGI GVHD	ELISA	High&moderate intensity	At least 2 agents including CNI	REG3a levels distinguished LGI GVHD from non-GVHD diarrhea (AUC: 0.80, PPV: 95%, NPV: 32%). REG3a levels (>151 ng/mL) at GVHD onset predicted non response to therapy at 4 weeks (3-fold higher), 1-year NRM (59% vs 34%), and 1-year OS (27% vs 48%) REG3a levels further risk-stratified patients who had either advanced clinical stage or histologic severity (34% vs 66% for 1 or 2 risk factors).
		S100	Reinhardt et al66	Human (52)	At onset or progression of GVHD	Calprotectin assay&sandwi ch-ELISA	MAC&RIC	CNI, MMF, MTX, ATG	S100A8/S100A9 and S100A12 levels were significantly increased in stool and serum of patients with GI aGVHD and extensive cGVHD compared with patients without GVHD. No correlation with the grade could be detected.
			Rodriguez-Otero et al67	Human (72)	At onset of GVHD	Sandwich-ELISA	MAC&RIC	CSA+ MMF CSA+ MTX	Fecal S100 concentrations were increased in stage II–III GI aGVHD but not stage I causing low sensitivity of the marker 31% (90% specificity) A high fecal S100 levels (≥ 100 μg/g) was strongly associated with SR-GVHD (CI of SR-GVHD: 93% vs 33%). S100 ≥ 100 μg/g independently correlated with a lower probability of CR (HR: 0.47)
			Chiusolo et al68	Human (59)	At onset of GVHD	ELISA	Not specified	Not specified	Fecal S100 was higher in patients with aGVHD than in non-GI-aGVHD. Levels increase with disease severity (with arbitrary cut-off of 160 mg/kg: Sen: 100%, Spec: 81.8%, PPV: 86%, NPV: 100%, AUC: 0.942)
	Liver-specific	TIM-3	Oikawa et al74	Murine	N/A	Flow cytometry	N/A	N/A	TIM-3 is up-regulated in aGVHD. Anti-TIM-3 antibody increases the severity of aGVHD
			Veenstra et al75	Murine	N/A	Same	N/A	N/A	TIM-3/gal-9 pathway acts as a suppressor of aGVHD
			Hansen et al76	Human (127)	At onset of aGVHD	Sandwich ELISA/Flow cytometry	MAC&RIC	CNI+MTX CNI+MMF	TIM-3 levels were significantly higher in patients with more severe mid-gut GVHD, compared with those with upper-gut GVHD, without GVHD, and normal control (medians: 11,550 vs 4670 vs 2710 vs 2285 pg/mL respectively). The levels were higher in samples collected closest to GVHD onset compared with earlier samples. The levels correlated with severity. Samples from patients with common diseases didn’t yield increased levels of TIM-3
		CK-18	Luft et al81	Human (55)	At the time of maximum GVHD	ELISA	MAC&RIC	Not specified	Both intestinal and hepatic GVHD were consistently associated with significant elevations of CK18F levels over baseline (5.6-fold change) CK18F levels decreased in responsive GVHD to immunosuppressive therapy and persisted in resistant GVHD Conditions that might represent relevant differential diagnoses (toxic mucositis, noncomplicated, infection-related diarrhea, and VOD) were not associated with CK18F elevation.
			Luft et al82	Human (48)	Longitudinally: before HCT, at escalation of immunosuppression, late	ELISA	MAC&RIC	CNI/MMF CNI/MTX	Significantly higher CK18F levels were detected in steroid-refractory GVHD indicating the ongoing epithelial death of target organs.
		HGF	Okamoto et al97	Human (38)	Serially from D0–120 post HCT	ELISA	Not specified	Not specified	HGF levels were increased in patients with than without aGVHD. The levels correlated with the grade of the disease.
			Paczesny et al88	Human (424)	D0–7–14–21–28–56–100 and at onset of GVHD	ELISA	MAC&RIC	FK506+MTX M CNI/MMF	HGF as part of a panel (along with IL-2Ra, TNFR1, IL-8) discriminated patients with and without aGVHD (AUC: 0.86). The panel predicted survival independently of GVHD severity. HGF was higher in visceral GVHD compared to skin-only GVHD.
			Harris et al62	Human (954) Multi-center	At onset of GVHD	ELISA	High&moderate intensity	CNI/MTX MF+MTX	HGF concentrations were elevated in LGI GVHD compared with isolated skin GVHD HGF levels perform poorly as a diagnostic biomarker distinguishing LGI GVHD from non-GVHD diarrhea (AUC: 0.60) and liver GVHD from other causes of hyperbilirubinemia (AUC: 0.59) High HGF concentrations correlated with significantly higher 1-year NRM
	Skin-specific	Elafin	Paczesny91	Human (492)	At onset of skin GVHD	ELISA	MAC&RIC	CNI+ another agent	Plasma elafin levels were twice as high in patients with skin GVHD compared to other groups (No GVHD, GI GVHD, Non-GVHD rash). Elafin levels at the time of diagnosis of skin GVHD correlated with maximum overall GVHD grade. Elafin is the best single discriminator for the diagnosis of GVHD in BMT patients with a rash compared to other biomarkers (AUC: 0.77) The 1-year NRM was more than double in the high elafin group (≥ 6000 pg/ml) compared with low elafin group (28% vs 11%) High elafin level at the time of GVHD diagnosis was significantly associated with a greater risk of death (hazard ratio of 1.78)
			Chacon98	N/A	N/A	N/A	N/A	N/A	There is immunohistochemical overexpression of elafin in both skin aGVHD and engraftment syndrome (ES). However, serum elafin hasn’t been reported to be elevated in ES

Abbreviations:

CSA: cyclosporine, MMF: myecophenolate mofetil, MTX: methotrexate, CNI: calcineurin inhibitors, FK506: tacrolimus.

MAC: myeloablative conditioning, RIC: reduced-intensity conditioning.

RT-PCR: reverse transcription polymerase chain reaction. ELISA: enzyme-linked immunosorbent assay.

Sen: sensitivity. Spec: specificity. PPV: positive predictive value, NPV: negative predictive value. HR: hazard ratio. NRI: net reclassification index. CI: cumulative incidence. AUC: area under the curve

CBT: cord blood transplantation.

VOD: veno-occlusive disease

SR-GVHD: steroid resistant GVHD. CR: complete response

Systemic biomarkers

MicroRNAs

MicroRNAs (miRNAs) are a class of small noncoding RNA (21- to 25-nucleotides) that negatively and positively regulate gene expression by translational repression or via induction in alterations in messenger RNA stability. miRNAs regulate gene function in a variety of ways and at multiple levels, particularly transcription, translation, and protein degradation[27]. Circulating miRNAs have been studied as novel, noninvasive biomarkers for many diseases such as cancer, sepsis, cardiovascular disease, liver injury, organ transplant rejection, and diabetes[28]–[31].

Researchers have conducted animal and human studies to investigate the role of miRNAs in the pathogenesis of aGvHD, and to screen for promising biomarkers and therapeutic targets [32]–[35]. Studies have found some lymphocyte-related serum miRNAs such as MiR-155 to be significantly up-regulated in aGvHD with a correlation between the level of serum MiR-155 and the severity of the disease[33]. However, MiR-155 is known to play a key role in inflammation regulation and immune response[36] and is unlikely to be specific for aGvHD. Furthermore, the expression of MiR-155 was the lowest among other miRNAs measured in the serum of patients with aGvHD[35]. Allo-HCT patients with high levels of another MiRNA, MiR-586, at day 7 were found to be at high risk of developing aGvHD[34]. However, levels of MiR-586 were easily affected by infections that are common at day 7 after allo-HCT. The lack of specificity of these studied MiRNAs has led to the simultaneous use of several MiRNAs to increase the specificity and diagnostic performance of these biomarkers.

Plasma miRNA signature has been studied as a biomarker for aGvHD. Xiao et al showed that evaluation of a panel of four microRNAs (miR-423, miR-199a-3p, miR-93, miR-377) was able to distinguish patients who developed aGvHD from those who did not (AUC: 0.80), and to predict the severity of the disease [35]. High expression of the miRNA panel was associated with poor overall survival. More importantly, elevated miRNAs can be detected at a median of 16 days before the diagnosis of aGvHD and were found to predict for the development of aGvHD when compared with another biomarker, sIL2Rα. Furthermore, this miRNA signature for aGvHD, unlike other studies of other MiRNAs or polypeptides, was not seen in the plasma of lung transplant recipients undergoing rejection, non-transplanted sepsis patients or patients with veno-occlusive disease[35]. This suggests a higher specificity of this miRNA signature for aGvHD. Serum or plasma miRNAs are highly stable in human peripheral blood[37] and are tissue-specific[38], making these biomarkers excellent targets for studying. There are other advantages of a miRNA signature including its ease of measurement, simplicity and lack of expense. Organ-specific association studies of miRNAs in the setting of aGvHD are lacking and could add to our improved stratification of patients to organ specific treatment strategies.

ST2

ST2 (suppression of tumorigenicity 2) is a member of the interleukin (IL)-1 receptor family and specifically binds to IL-33. ST2 is present in 2 isoforms: a membrane form expressed on hematopoietic cells, specifically T helper 2 (Th2) cells, and a soluble form secreted by endothelial and epithelial cells in response to inflammatory injury [39], [40]. Soluble ST2 appears to act as a receptor for IL-33 limiting its access to Th2 cells and promoting the Th1 phenotype, which has been associated in aGvHD pathophysiology[41], [42].

In a large two center study of 673 recipients of myeloablative or nonmyeloablative HCT, patients with high ST2 measured as early as 14 days after transplant had an increased risk of non-relapse mortality (NRM)[43]. Moreover, ST2 level predicted the response of aGvHD to treatment. Patients with high ST2 levels (defined as levels greater than 50% above the responder’s median value) were less likely to respond to treatment, while those with low levels were more likely to respond to therapy even if they had high-grade aGvHD using clinical criteria. The addition of ST2 to the clinical risk factors has reclassified almost 60% of patients receiving all types of conditioning regimens. Furthermore, a strong additive effect was noted when including the concentrations of both ST2 and REG3a (see below) to the clinical characteristics in order to improve the accuracy of the risk stratification. Similar results were obtained when testing the efficacy of ST2 in risk-stratifying Cord Blood Transplantation (CBT) patients[44].

Despite the promising results of these studies in demonstrating a particular significance of ST2 measurements in post-transplant mortality risk stratification, many questions remain. ST2 concentrations seem to vary significantly across conditioning intensities, necessitating separate thresholds for each conditioning intensity to be identified. Furthermore, the timing for measuring ST2 was arbitrarily chosen in these studies, and prospective evaluation of serial ST2 measurements is needed to refine the predictive value of this biomarker.

Biomarkers of Immune Activation

The persistent activation of the immune system that occurs early after donor graft infusion leads to excessive cytokine production (cytokine storm). Cytokines are important to the pathogenesis of aGvHD[45]. Many studies have examined cytokines and their receptors as potential aGvHD biomarkers. Herein we review studies that examined the diagnostic and prognostic utility of serum levels of TNF receptor 1, IL-7, and sBAFF.

TNF receptor 1

TNF-α is an inflammatory cytokine frequently implicated in the pathogenesis of aGvHD. It is the most frequently reported immune activation marker to be elevated prior to GvHD onset[46]. Researchers have measured TNF receptor 1 (TNFR1) as a surrogate marker for TNF-α. Increased levels of TNFR1 at day 7 equal or greater than 2.5 x baseline prior to transplant significantly correlated with the eventual development of severe aGvHD and with treatment related mortality in several studies[47–49]. Interestingly, day 7 serum solubleTNFR1 levels were not significantly associated with grade II–IV aGvHD in these studies, suggesting that TNFR1 day7/baseline ratio has better predictive value. Another important observation of these studies is the low sensitivity of TNFR1 ratio (equal to or greater than 2.5 x baseline) in predicting severe aGvHD (sensitivity <40%) since the majority of patients who develop grades II–IV aGvHD have a TNFR1 day7/baseline ratio less than 2.5. Nevertheless, in patients with a TNFR1 ratio more than or equal to 2.5 x baseline, the likelihood of developing significant aGvHD is adequately high (~60%) to justify the future use of TNFR1 ratio-based preemptive treatment strategies. The screening performance of TNFR1 ratio may be strengthened by combining it to other biomarkers as a composite panel (See Table 2) (AUC=0.77 and NPV= 0.87)[50]

Table 2.

Examples of biomarker panels that combine several markers of aGVHD to increase the specificity or predictive/diagnostic power.

Panel biomarkers	Study	Type/N	Time	Assay	Conditioning regimen	GVHD Prophylaxis	Outcome
TIM3, IL6, sTNFR1, ST2	McDonald et al77	Single center (317)	D 7 through d 70 after HCT in cohort 1&2 (with GI GVHD) D 14±3 after HCT in patients without GVHD	ELISA	MAC	CNI+MTX CNI+MMF	The panel predicted development of grade 3–4 GVHD at median of 4 days before initiation of therapy (AUC: 0.88). TIM3 predicted subsequent grade 3–4 GVHD (AUC: 0.76). Plasma ST2 and sTNFR1 predicted 1-year NRM (AUC: 0.90).
IL-2Ra, TNFR1, HGF, IL-8, elafin, REG3a	Levine et al92	Multicenter (112)	Onset of GVHD D 0, 14, 28 after therapy initiation	ELISA	MAC/RIC	Not specified	D 0 and 14 biomarker panel independently predicted nonresponse at d 28 (OR: 2.98, OR: 6.32). D 0 biomarker panel independently predicted mortality by day 180 (OR: 4.61, AUC: 0.7210). D 28 biomarker panel predicted d 180 NRM (OR: 7.43).
Several soluble and cellular biomarkers	Te Boome et al94	Single center (48)	before and at predetermined time points after first MSC infusion	ELISA	MAC/RIC	Not specified	The panel was predictive for mortality (HR 2.924) when measured before MSC-administration. ST2 was only predictive for mortality 2 weeks after but not before MSC-administration (HR 2.389).
TNFR1, ST2, REG3a	Levine et al93	Multicenter (492)	48 h before or after therapy initiation	ELISA	MAC/RIC	CNI+MTX CNI+MMF CNI+Sirolimus Post-HCT cyclophosphamide	The CI of 12-month NRM significantly increased as GVHD score increased (8% vs 27% vs 46% for score 1, 2, 3). The response rates to primary GVHD therapy decreased as the GVHD score increased (86% vs 67% vs 46% for score 1, 2, 3). The development of GI GVHD in patients who presented with skin GVHD only increased as the the score increased (19% vs 29% vs 34% for score 1, 2, 3). First biomarker-based score that can be used to guide risk-adapted therapy.

Abbreviations:

MMF: myecophenolate mofetil, MTX: methotrexate, CNI: calcineurin inhibitors.

MAC: myeloablative conditioning, RIC: reduced-intensity conditioning.

RT-PCR: reverse transcription polymerase chain reaction. ELISA: enzyme-linked immunosorbent assay.

MSC: mesenchymal stromal cells

CI: cumulative incidence. HR: hazard ratio. AUC: area under the curve.

IL-7

Interleukin-7 (IL-7), is a growth factor that represents the principal homeostatic cytokine for T cells and is important for B and T-cell maturation. IL-7 promotes immune reconstitution after allogeneic HSCT and is needed for the development of acute GvHD in murine models[51], [52]. High levels of IL-7 along with other cytokines leads to homeostatic peripheral expansion of donor T cells within 30 days post-transplant. Given this key role in initial T-cell recovery, many groups have measured the levels of IL-7 over sequential time points before and after allo-HCT. In small studies, IL-7 levels at D+14 predicted the subsequent development of aGvHD (PPV: 85%, NPV: 88%) and higher levels were strongly associated with more severe grades of aGvHD[53]. By applying a multivariate model, high serum levels of IL-7 by day 14 after transplantation was the factor most strongly associated with the probability of developing grades II–IV aGvHD both in myeolablative[54] and reduced intensity conditioning[55]. It is important to note, however, that some other studies have reported less TRM in patients who recover T cells more rapidly[56] suggesting that the ratio among the subsets of the lymphocytes rather than the size determines the aGvHD developments and the outcomes. The predictive value of IL-7 levels still requires support from larger, prospective trials.

sBAFF

B cell activating factor, also known as B cell survival and activation factor, acts as a potent B-cell activator and it has been shown to play a crucial role in the proliferation and differentiation of these cells in mice and humans. It has been linked to a variety of autoimmune diseases. sBAFF is required for the reconstitution of B cells after myeloablation in animal models. Cho et al found that elevated sBAFF levels at any time point in the early days post-HCT were associated with decreased risk of developing aGvHD[57]. These findings suggest not only a predictive role of sBAFF during the peri-transplantation period, but also suggest that elevated levels may even confer protection against the development of aGvHD. This is to be contrasted with the notable recent reports that demonstrated an association between high sBAFF levels and the occurrence of chronic GvHD (cGvHD)[58]. This highlights the difference in pathogenesis between aGvHD, which is mediated mainly by donor T cells, and cGvHD where donor B cells play an important role in the pathophysiology of cGvHD. Despite the recent reports that have linked B cells to aGvHD[59], the role of B cells in the pathogenesis of aGvHD remains uncertain. Furthermore, T cells have been shown to express BAFF receptors[60]. Therefore, there could be a role for sBAFF in the regulation of T-cell dependent immunity. Further experimental and clinical studies are needed to gain more insight into the roles of BAFF and B cells in the pathogenesis of aGvHD

The use of sBAFF as a biomarker for aGVDH prediction may be complicated by a number of factors. First, BAFF levels are increased in the setting of B lymphopenia. Furthermore, the levels are also affected by medications, particularly high-dose steroids that are commonly used in aGvHD. Finally, current ELISA assays underestimate BAFF levels and precise quantification of the biomarker is lacking.

Organ-specific biomarkers

Lower Gastrointestinal Tract (LGI)- specific Biomarkers

LGI GvHD occurs in up to 60% of patients following allo-HCT. The involvement of the lower GI tract with aGvHD is often severe, and is characterized by voluminous secretory diarrhea, with or without hematochezia, and abdominal cramps. The etiology of diarrhea following allo-HCT is a common diagnostic dilemma often indistinguishable from other causes of diarrhea (conditioning chemotherapy related, administration of antibiotics, clostridium difficile-associated diarrhea, or cytomegalovirus infection). The gastrointestinal involvement of aGvHD remains a major cause of morbidity; and the use of biomarkers to predict the occurrence or estimate the maximal severity would be helpful. The most studied LGI-specific biomarkers are: REG3α, CK-18, S100, and TIM-3.

REG3a

REG3α (regenerating islet-derived 3-α) is an antimicrobial protein expressed in Paneth cells and secreted into the crypt microenvironment. REG3a may play a protective effect for intestinal stem cells (ISCs), which are important cellular targets of GvHD in the GI tract[61].

REG3α has been shown to have a very good diagnostic and prognostic performance in multicenter studies with large number of patients. REG3a distinguishes LGI GvHD from non-GvHD diarrhea with excellent positive predictive value and AUC (95% and 0.8 respectively). REG3α was the best diagnostic biomarker for LGI GvHD and additional biomarkers (HGF and CK18) as a composite panel provided minimal increased sensitivity or specificity. Furthermore, the levels of REG3a at onset of LGI GvHD predicts the response to first-line therapy and 1-year NRM[62], [63].

REG3α, along with clinical severity and histologic severity, provides important prognostic information before the initiation of therapy rather than at the time of maximum grade of GvHD. The three factors (REG3α levels, clinical and histologic severity) independently predict the lack of response to therapy and likely integrated into single grading system that will permit better risk stratification of patients with severe LGI aGvHD.

S100

S100 are pro-inflammatory proteins that play an important role in many inflammatory disorders, such as inflammatory bowel diseases (IBD) and rheumatoid arthritis RA[64], [65]. S100 proteins are released by activated or damaged phagocytes under conditions of cell stress during infections and autoimmune disease and they represent promising novel therapeutic targets. Studies have attempted to evaluate the diagnostic and prognostic value of serum and fecal levels of some of S100 proteins (S100A12 and calprotectin: S100A8/S100A9) in the setting of LGI aGvHD[66]–[68]

The major advantage of these markers was not the capability to diagnose GI-GvHD but rather their ability to predict the response to steroids, and therefore risk-stratifying these patients. The sensitivity of fecal S100 proteins for the diagnosis of stage 1 GI aGvHD was quite low (with no sensitivity improvement with cutoff decrease) and was unable to discriminate aGvHD from other causes of diarrhea. However, the S100 levels were predictive for responses to treatment. Furthermore, fecal S100 (along with other fecal markers: α1-antitrypsin and elastase) were particularly useful re-stratifying patients with stage 2 GI GvHD, allowing patients with the worst prognosis to be identified. Interestingly, the probability of steroid-resistant GI aGvHD was as high as 100% if two fecal markers were high and as low as 0% if the 2 markers were low.

Fecal proteins represent attractive noninvasive biomarkers to evaluate LGI aGvHD. The biomarkers are immediately available and can be easily collected. Moreover, if validated they they may represent a better tool to grade, risk-stratify and monitor the response to treatment in patients with LGI aGvHD compared to histologically variable GI biopsies and measurement of stool volumes.

TIM-3

The TIM (T-cell immunoglobulin domain and mucin domain) family of genes was first described in 2001[69]. The critical role of these genes products (3 proteins: TIM-1, -3, and -4) in regulating immunity is beginning to emerge[70]. TIM-3 was the first described and has been the most studied member of the family[71]. To date, the in vivo functions of TIM-3 and its role in immunoregulation remained largely unknown with discrepancy between studies. Interactions between TIM-3 and its ligand galectin-9 (gal-9) play a role in autoimmune disorders, chronic infection, tumor immunity, and transplantation[72], [73]. In murine models of aGvHD, TIM-3 has been shown to be up-regulated with a possible inhibitory role to the interaction between TIM-3 and its ligand in the pathogenesis of aGvHD[74], [75]

The predictive value of serum TIM-3 levels in the setting of GvHD has been studied by Hansen et al[76]. Although TIM-3 levels correlated with the occurrence and severity of the disease, the area under the receiver operating characteristic (ROC) curve was significant only for severe mid-gut GvHD compared to all other aGvHDs including upper GI aGVHD (0.79 vs 0.59). These findings are inconsistent with the recent paper from McDonald et al[77], where the measurement of TIM-3 was the most useful analyte among others for predicting grade 3–4 vs grade 0–2 aGvHD (AUC: 0.76) (See Table 2).

CK-18

Cytokeratin-18 (CK18) is an intermediate filament that is typically expressed in the epithelia of the digestive, respiratory and urogenital tracts[78]. The initiation of the apoptotic process results in activation of certain caspases which cleave a variety of cellular substrate including CK18. Cleavage at a particular site (DALD-S) gives rise to a neo-epitope termed cytokeratin-18 fragments (CK18Fs) which are released into the serum [79]. Intra-cryptal apoptosis is the histopathologic hallmark in GvHD and is used in the grading of GvHD[80]. Based on the hypothesis that apoptosis at the level of target organs in aGvHD leads to increase in CK18F and that CK18F mirrors the pathogenetic end point of GvHD, researchers have investigated whether CK18F provides a tool for sensitive assessment of GvHD-associated apoptotic activity and grading of the disease[81], [82]. Although CK18F levels correlated with GI and hepatic GvHD as well as response to immunosuppressive therapy, these studies were not designed to systematically determine the diagnostic features of this biomarker (sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV)). Apoptosis is not GvHD-specific and more studies are needed to validate the diagnostic performance of its associated biomarkers in the setting of aGvHD.

Liver-specific Biomarkers

HGF

Hepatocyte growth factor (HGF) is a cellular growth factor that targets epithelial and endothelial cells including hepatocytes, by activating a tyrosine kinase signaling cascade after binding to proto-oncogene c-Met receptors[83], [84]. Serum HGF concentrations have been reported to be elevated in liver disease, chronic renal failure, systemic inflammatory response syndrome (SIRS), and malignant diseases such as leukemia, breast cancer, and gastric cancer[85]–[87].

Although the levels of HGF were found to be elevated in patients with visceral GvHD, HGF performs poorly as a diagnostic biomarker. HGF failed to distinguish LGI aGvHD from non-GvHD diarrhea (AUC: 0.6)[62]. Likewise, HGF concentrations are elevated in liver GvHD compared with asymptomatic patients but are comparable to concentrations in patients with non-GvHD hyperbilirubinemia (AUC:0.59)[62]. A small number of patients develop hyperbilirubinemia caused by non-GvHD transplantation-related complications such as veno-occlusive disease (VOD) and these complications can be discriminated clinically from GvHD. In spite of this, a biomarker(s) that differentiates liver aGvHD from these complications would be clinically meaningful. HGF seems to perform significantly better as a diagnostic biomarker when integrated in a composite biomarker panel[88]. (See Table 1)

In spite of the poor diagnostic capabilities of HGF, this biomarker has a significantly better prognostic performance. The biomarker panels that involved HGF were shown to predict 1-year NRM and long-term survival independently of GvHD severity, and HGF was the primary contributor to the prognostic significance of these panels[62], [88]

Skin-specific Biomarkers

Elafin

Skin-derived antileukoproteinase (SKALP), otherwise known as Elafin, is an elastase-specific protease inhibitor expressed mainly in epithelial cells. Elafin is not present in normal keratinocytes but overexpressed in a variety of inflammatory skin disorders such as psoriasis[89]. In vivo studies have shown that inflammatory cytokines that are secreted in patients with aGvHD can stimulate the expression of elafin in epidermal cells [90]. Immunohistochemistry studies revealed an overexpression of elafin in GvHD skin biopsies. Elafin is expressed by the target epidermal cells rather than the effector cells that injure all three-target organs (GI, Liver, skin), which explains the higher specificity of elafin as a biomarker for skin aGvHD. Although produced locally as an antiprotease secreted in response to cytokines, elafin is readily detected in the systemic circulation as well.

When the diagnostic features of serum elafin as a biomarker of skin aGvHD were compared to other biomarkers of aGvHD (TNFR1, IL2Rα, IL8, HGF) (See Table 1), elafin was found to be the best single discriminator for the diagnosis of skin GvHD from other etiologies of rash (drug rash, engraftment syndrome, leukemia cutis, and viral or fungal rash). The area under the curve (AUC) of elafin was the highest (0.77) compared to the above biomarkers. Furthermore, elafin correlated with the maximum overall GvHD grade and 1-year NRM, making its prognostic value appealing as well[91].

Biomarker Panel

When several clinically-useful biomarkers are present, typically no particular one is satisfactory alone in terms of sensitivity or specificity for the diagnosis or prediction of a specific disease. Researchers have identified potential biomarker panels for aGVHD either by combining individual biomarkers that have been studied previously or by performing discovery studies that compare samples from patients with severe aGVHD with samples from patients without aGVHD. By analyzing these samples, biomarkers that provide the best discrimination between the two groups are identified. Afterwards, the panel will be developed, validated and tested for operating characteristics. Summary of biomarker panels is presented in Table 2.

Recently, many groups have used biomarker panels to identify patients at the extremes of outcomes. These panels have been shown to provide prognostic and predictive information of aGvHD outcomes, including maximal severity, response to therapy, and mortality[77], [92]. The predictive values of these panels were independent and additive of the clinical status of the patient (onset grade, unrelated vs related donor and peripheral vs bone marrow vs cord as a source of stem cells) with very good area under the curve(0.7–0.9) (Table 2). However, the positive predictive values of the panels remained in the range of 40–50%, indicating that the false positives would still equal or outnumber the true positives. Recent reiteration of the biomarker panel by the University of Michigan group (Ann Arbor GvHD score) examined the utility of measuring 3 biomarkers (TNFR1, REG3α, and ST2) at the time of aGvHD diagnosis to create a model algorithm that predicts NRM 6 months later in a well-powered study[93]. The algorithm defined three distinct scores whose risk of NRM increased with each increasing grade (independently of other risk factors, such as donor type, age, conditioning regimen, and HLA match) and may prove useful in the guidance of aGvHD therapy. The value of biomarkers in predicting the risk of complications and mortality following HCT will be studied in BMT-CTN Protocol 1202.

Although the integration of a biomarker panel in clinical practice still needs further investigation, these panels may also provide a valuable alternative or addition to composite end points in clinical trials evaluating novel therapies for aGvHD and improve the success rate of these trials. Biomarker panels may provide an opportunity for a better selection of patients who are likely to benefit from expensive or potentially risky therapies tested in clinical trials as well as a close monitoring of response. To this end, Te Boome et al evaluated the utility of biomarkers in the prediction of either resolution of aGvHD or survival in patients with steroid-refractory aGvHD treated with mesenchymal stromal cells (MSCs) in a prospective phase II trial[94]. Surprisingly, biomarkers that were previously reported to be predictive of outcomes (ST2[43] and Levine panel[92]) failed to correlate with clinical severity or predict resolution of aGvHD, likely due to the relatively low number of patients.

These studies represent the first demonstrations of how GVHD biomarkers panels and biomarker-based scores may be potentially incorporated into clinical care.

Summary

Many flaws exist in the current clinically-based grading systems of aGvHD. Maximal clinical severity, which grading tools rely on, can only be assigned retrospectively after the response to treatment is known. Withholding treatment for GvHD or waiting for 14–28 days to evaluate for refractoriness to therapy may lead to extensive and irreversible damage, especially in the GI tract, resulting in unacceptable high rates of morbidity and mortality. If it were possible to predict the ultimate severity of aGvHD before or at the onset of symptoms, preemptive therapy (e.g. anti-T cell therapy) could be given to blunt the intensity of tissue damage. Whether this approach, even when implemented early will work to change the natural history of GvHD remains to be seen and proven in prospective clinical trials. Even patients who respond to low doses of steroids can experience unnecessary infectious complications as well as other morbidities such as diabetes mellitus and avascular necrosis.

Previous studies have clearly established correlation between individual as well as combined biomarkers and outcomes of aGvHD. However, inconsistent results among different clinical centers and studies are noted and mentioned in this review. This inconsistency can be attributed to the high heterogeneity of patient groups receiving various treatments as well as the variation in study protocols and treatments. Despite the statistically significant operating characteristics of many of the biomarkers discussed and the composite panels, the use of these data in clinical practice should currently be reserved for use in well designed prospective clinical trials. Although the data derived from biomarker-based scoring systems predicts outcomes better than clinical scoring systems, the positive predictive value of these biomarkers is still low enough and is most useful for patients who score at either end to justify preemptive therapy. Moreover, reconciling the results of the various studies done at different centers to examine the value of different biomarkers in the prediction of aGvHD outcomes will be difficult due to the differences in timing, intensity of conditioning therapy, choice of GvHD prophylaxis, source of hematopoietic cells, and statistical methods. This suggests that predictive biomarkers might need a coordinated multicenter approach with coordinated times of sampling and centralized analysis of biomarker levels. Another attractive approach would be to incorporate the clinical characteristics and risk factors with the biomarker scores (such as the Ann Arbor score) to improve the predictive power of these scores and their treatment algorithms. Biomarkers represent a compelling yet still unproven tool for reproducible prediction of aGvHD, the severity of aGvHD and the outcome of therapy.

Highlights.

• Biomarkers predict the severity and outcomes of aGVHD better than clinical grading systems.

• Integrating clinical characteristics into biomarker scores may improve the predictive power of both.

• Multicenter studies are needed to validate these biomarkers in different settings.