Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a curative therapy for many patients with hematologic malignancies and non-malignant diseases, but continues to be associated with significant mortality (1–3). Chronic graft-versus-host disease (cGVHD) remains the leading cause of long-term morbidity and mortality after allo-HSCT and occurs in 30 – 70% of the transplanted adults or children who survive longer than 100 days (4, 5). Especially, the search for validated biomarkers is of utmost importance (6). In 2014, the National Institute of Health (NIH) updated its biomarker report of 2005 and established a standardized nomenclature for three important groups of biomarkers in cGVHD: 1.) Diagnostic biomarkers that are intended to facilitate the diagnosis of cGVHD, 2.) prognostic biomarkers that can predict disease onset and 3.) predictive biomarkers, used to predict the therapeutic outcome of the patients (7).

In acute GVHD (aGVHD), all three biomarker groups are already well studied, and biomarker panels to facilitate the prognosis and diagnosis of aGVHD are established (8–13). The pathogenesis of chronic disease, by contrast, is much more heterogeneous and can affect nearly every organ, complicating the search for reliable biomarkers (14). Despite advances, the incidence of cGVHD after allo-HSCT remains remarkable, highlighting the urgent need for reliable biomarkers to identify at-risk patients and guide pre-emptive therapies (6, 15). Given that a significant proportion of patients start treatment with already non-reversible manifestations, prognostic biomarkers identifying patients before highly morbid forms of cGVHD manifest and permitting pre-emptive treatment are urgently needed (16). In recent years, several studies have attempted to find diagnostic biomarkers for cGVHD. In this regard, CXCL9, CXCL10, and BAFF have been identified as promising markers; however, these studies have often yielded conflicting results (17–20). In order to classify prognostic biomarkers in cGVHD only a few studies have been conducted. Despite the undeniable clinical potential of this type of biomarker, no reliable single protein has been validated to date (21–25).

Moreover, many biomarker studies have focused on identifying markers without considering the prior history of GVHD. The separation of cGVHD in the two groups: 1.) de novo cGVHD (cGVHD without prior acute GVHD) and 2.) quiescent cGVHD (patients with prior acute GVHD and subsequent cGVHD) is not only important regarding the outcome of these patients (26). It may also be relevant in the search for prognostic biomarkers, as different immunological processes might be operating during the development of the disease in these two groups (22, 27).

Therefore, in this study, we focused on identifying potential prognostic biomarkers for cGVHD in patients after allo-HSCT and compared these biomarkers between the two different cGVHD subcohorts: de novo cGVHD and quiescent cGVHD. Thereby, we included biomarkers already identified in aGVHD as well as cGVHD and explored their prognostic potential in cGVHD. Moreover, we searched for new, promising candidates that had not been previously described.

This study aimed to identify prognostic biomarkers for subsequent cGVHD developing after aGVHD or de novo without preceding aGVHD. Therefore, patients were separated into four different cohorts: group 1.) neither aGVHD nor cGVHD (label “never GVHD”), group 2.) de novo cGVHD without prior aGVHD (de novo cGVHD), group 3.) aGVHD but no cGVHD (resolved aGVHD) and group 4.) quiescent cGVHD after previous aGVHD (quiescent cGVHD). Serum samples collected before cGVHD onset enable the identification of biomarkers that predict disease risk, underscoring their prognostic rather than diagnostic value. To ensure biomarkers were not specific to aGVHD, all forms of aGVHD (classic or late) had to be fully resolved at the time of sample collection, and prednisolone doses had to be below 0.5 mg/kg/day due to its known effect on sBAFF levels (18).

To date, the exploration of prognostic biomarkers in cGVHD has been limited by the disease’s heterogeneity. In particular, the influence of ongoing aGVHD and the impact of corticosteroid immunosuppression on cytokine levels have complicated the search for reproducible prognostic markers in previous studies. Moreover, analyzing all cGVHD patients as a single group in terms of disease onset may have prevented the discovery of validated biomarkers that might be specific for a subcohort of cGVHD which has been recently described in children (29). This exploratory study lays the groundwork for a comprehensive biomarker analysis that should include a larger, multicenter cohort and independent validation.

One of the strengths of our study is the establishment of a clear framework to exclude cytokine alterations solely related to aGVHD by omitting all patients with ongoing aGVHD at the time of sample collection. In addition, we minimized the confounding effects of corticosteroid therapy by including only patients receiving ≤0.5 mg/kg body weight/day prednisone equivalent at sampling, as glucocorticoids can substantially influence serum protein expression (18, 24). Previous studies have primarily compared patients who developed cGVHD with those who did not, without accounting for the potential impact of different subcohorts on biomarker levels (22). Moreover, studies investigating risk factors for cGVHD have already recognized the need to distinguish between the two clinical types of cGVHD (30, 31). In our analysis, we therefore added a separate analysis distinguishing patients who developed de novo cGVHD from those who developed quiescent cGVHD and compared each group with matched controls - that is, de novo cGVHD with patients who never developed GVHD, and quiescent cGVHD with patients who had resolved acute GVHD, reflecting the clinical situation.

Acute GVHD is a well-established risk factor for the subsequent development of cGVHD (32, 33). It has been proposed that tissue damage caused by acute inflammation or its treatment promotes the onset of quiescent cGVHD (34, 35). These assumptions, however, do not apply to de novo cGVHD, except in cases where it follows a very mild untreated episode of aGVHD. In such instances, the pathogenesis – and consequently the contributing biomarkers – of de novo and quiescent cGVHD would be expected to overlap. In contrast, our findings provide evidence against this notion. We identified distinct prognostic biomarker panels predicting de novo versus quiescent cGVHD, suggesting that these two clinical entities may arise through partially distinct pathogenic mechanisms. These insights may help elucidate the biological underpinnings of cGVHD development.

Whereas CXCL9 was previously described as a diagnostic biomarker for cGVHD (17, 36), our data now identify elevated CXCL9 levels as a prognostic biomarker for de novo cGVHD. CXCL9 is strongly induced by IFN-γ (37). Binding to its receptor CXCR3 promotes the recruitment of activated lymphocytes to the inflammatory tissue (38, 39). In our cohort, CXCL9 serum levels were significantly increased 90 days after allo-HSCT in patients who later developed de novo cGVHD compared with their matched controls who never experienced GVHD. These findings suggest that IFN-γ–driven processes may contribute to the early pathogenesis of de novo cGVHD. In contrast, CXCL9 did not emerge as a prognostic marker for quiescent cGVHD, likely because CXCL9 serum levels were already elevated in patients who had resolved aGVHD, irrespective of subsequent cGVHD development.

The BAFF system is a key regulator of B-cell survival, maturation, and activation. At day +90 after HSCT, serum BAFF levels did not differ between patient cohorts. This likely reflects a general early post-transplant increase in B-cell–modulating cytokines, supporting reconstitution of the B-cell compartment after myeloablative conditioning. Echoing previous observations in patients with active cGVHD (18), elevated BAFF levels six months after transplantation were high in patients with the subsequent development of cGVHD but also found elevated in patients with aGVHD only. Importantly, our data now extend the prognostic association specifically to de novo cGVHD.

A novel finding of this study is the identification of the soluble form of the receptor for advanced glycation end products (sRAGE) as a potential biomarker for cGVHD. Advanced glycation end products (AGEs) constitute a heterogeneous group of molecules that interact with their membrane-bound receptor RAGE, triggering oxidative stress, pro-inflammatory cytokine production, and tissue inflammation (40). The soluble receptor sRAGE acts as a decoy by binding circulating AGEs and thereby inhibiting AGE–RAGE signaling and its downstream inflammatory effects (41). In our cohort, sRAGE emerged as a potential prognostic biomarker for de novo cGVHD, with significantly lower serum levels compared to patients who never developed GVHD. In contrast, sRAGE levels in patients who later developed quiescent cGVHD did not differ from those in their matched controls with resolved aGVHD.

CCL4 is a potent chemoattractant for monocytes and T cells that exerts its effects through binding to its receptor CCR5 (42). While CCR5 blockade has been shown to ameliorate visceral GVHD (43), CCL4 itself has not yet been described as a biomarker for any form of GVHD (42). In our cohort, CCL4 serum levels were elevated in patients with prior aGVHD and were even higher in those who subsequently developed quiescent cGVHD. Notably, CCL4 also demonstrated predictive value for de novo cGVHD, as shown here for the first time. Together, these findings identify CCL4 as a promising new candidate for a prognostic biomarker for both forms of cGVHD.

For the first time, we describe interleukin-6 (IL-6) as a promising prognostic biomarker for quiescent cGVHD since this patient cohort had the highest levels of IL-6. One explanation is that subclinical inflammatory activity persisting after aGVHD, despite clinical resolution, is captured by elevated IL-6 levels. This suggests that IL-6 may serve as a promising prognostic biomarker for identifying patients at risk of quiescent cGVHD early after aGVHD.

Th17 cells and their key effector cytokine IL-17A are known mediators in the pathophysiology of cGVHD, promoting disease progression toward sclerotic manifestations (27, 44, 45). In this study, we report for the first time that elevated IL-17A serum levels precede the development of quiescent cGVHD.

High PAI-1 levels are indicative of endothelial damage and have been linked to various vascular disorders (46–48). Nürnberger et al. reported elevated PAI-1 concentrations in several post-transplant complications, including GVHD (49). In our study, PAI-1 emerged as the most consistent prognostic biomarker for quiescent cGVHD, with significantly higher levels at both time points compared with patients who experienced but resolved aGVHD. Given that aGVHD causes substantial endothelial and tissue injury (27), we hypothesize that patients who do not progress to cGVHD can fully repair this damage, whereas those who develop quiescent cGVHD may fail to achieve complete endothelial recovery, leading to chronic disease (27).

The soluble chemokine CX3CL1, previously linked to tissue damage after conditioning in aGVHD showed elevated serum levels in patients who developed quiescent cGVHD compared to those with resolved aGVHD (27, 50). CX3CL1 levels were also higher in quiescent than in de novo cGVHD, supporting a mechanism driven by prior tissue injury.

Another pro-inflammatory chemokine, CXCL1, is a potent chemoattractant for neutrophils during inflammation (51). We find a potential in CXCL1 to serve as a prognostic biomarker for the development of quiescent cGVHD.

Interleukin-10 (IL-10) is a key anti-inflammatory cytokine. Ninety days after transplantation, elevated IL-10 levels were observed only in patients with resolved aGVHD, likely reflecting residual anti-inflammatory responses. At day +180, IL-10 was prognostically increased in patients who later developed quiescent cGVHD despite high inflammatory activity. Overall, IL-10 serum levels remained low and warrant further investigation.

Although our study identified several potential biomarkers with in part different performance depending on the type of onset of cGVHD, certain limitations should be acknowledged. Biomarker levels were not consistent across the two sampling time points, underscoring the need for longitudinal studies with more frequent sampling. The cohort size was limited, and larger patient populations will be required to replicate the findings. In addition, a larger patient population may reveal correlations between biomarker levels and cGVHD severity. In our cohort, we could not find such correlations (data not shown). We have not applied multivariate analysis as sample sizes were limited and therefore the identified biomarkers may not have independent prognostic information. A time-to-event framework would assess subsequent cGVHD risk and should be applied in future studies with larger patient numbers. Finally, this study was not designed to assess associations between biomarker levels and specific organ manifestations, which may influence serum protein profiles.

In conclusion, our study addresses the challenges of identifying prognostic biomarkers for cGVHD by applying a rigorous framework to minimize confounding factors. We identified several promising biomarkers for de novo cGVHD – particularly CXCL9, BAFF, and sRAGE – with strong predictive potential. In contrast, IL-6, IL-17A, PAI-1, CX3CL1, CXCL1, and IL-10 emerged as notable prognostic indicators for quiescent cGVHD, reflecting distinct underlying inflammatory processes. CCL4 was identified as a promising biomarker common to both subtypes. Together, these findings advance efforts toward more personalized prediction and management of cGVHD.