Oxidative DNA damage in reconstituting T cells is associated with relapse and inferior survival after allo-SCT

Allogeneic hematopoietic stem cell transplantation (allo-SCT) is the only curative treatment option for several hematologic malignancies, including myelodysplastic syndrome or acute myeloid leukemia (AML), with high-risk features or insufficient responses toward conventional therapy.¹ The success of allo-SCT is owed to the activity of the donor’s immune system, in particular, the T cells against residual malignant cells, the so-called graft-versus-leukemia (GVL) effect.² However, donor cells can also cause a multisystem immune disorder termed as graft-versus-host disease (GVHD), which is a major cause of morbidity and mortality and affects 40% to 60% of the patients.³ 

Because of the central role of the (transplanted) immune system in mediating allo-SCT’s beneficial effects, it has been possible to reduce the intensity of preparatory conditioning protocols and use, among other strategies, the tapering of prophylactic immunosuppression (for GVHD) and/or donor lymphocyte infusions to treat relapses.⁴ Unfortunately, relapse of the original disease remains the primary cause of death, and a better understanding of the underlying mechanisms is required to improve the outcome. To date, several nonexclusive mechanisms that promote (allo-) immune escape of malignant cells have been described.⁵ Under selective pressure, malignant cells can lose HLA expression. In fact, it has been reported that relapsing AML blasts after allo-SCT display lower HLA expression than that at the time of initial diagnosis.⁶ Naturally, numerous studies have focused on T cells as the main effectors of antitumor immunity in allo-SCT: expression of immune checkpoint ligands (eg, programmed cell death protein 1 [PD-1]⁷), signs of exhaustion,⁸ and the loss of proinflammatory cytokine production (eg, interferon gamma [IFN-γ]) have emerged as drivers for disease relapse and/or outgrowth. In a recent seminal study, Zeiser and colleagues⁹ showed that relapsing AML blasts after allo-SCT perform aerobic glycolysis at high rates and release lactic acid. This bioactive metabolite is then taken up by T cells, leading to bioenergetic reprogramming that favors suboptimal GVL activity. In fact, lactic acid accumulation as a promoter of immune escape has been previously described in solid malignancies and belongs to the group of tolerance-promoting metabolites that can accumulate within the tumor microenvironment.¹⁰ Reactive oxygen species (ROS), such as hydrogen peroxide, are further prime examples from this metabolite class. Abundance of short-lived and hyperreactive ROS yields a pathologic condition called oxidative stress. In fact, oxidative stress represents a hallmark of cancer (eg, in chronic lymphocytic leukemia¹¹) that exerts detrimental effects on T cells by downregulating the CD3ζ chain, promoting apoptosis, and providing a survival advantage for suppressive regulatory T cells (T_Regs) over conventional T cells.^12-14 One of the underlying pathogenic mechanisms is oxidative DNA damage, which may lead to strand breaks, mutations, epigenetic instability, and chromosome abnormalities.¹⁵ In this context, 8-hydroxydeoxyguanosine (8-OHdG) represents the most frequent oxidative DNA damage marker, with cellular 8-OHdG levels correlating with systemic oxidative stress levels.¹⁶ However, a well-balanced ROS production by mitochondria is required for proper T-cell function and activation.¹⁷ Here, ROS act as an important second messenger that promotes continued T-cell receptor signaling.

Oxidative stress after allo-SCT has been reported in preclinical models and patients.^18,19 Reconstituting T cells in patients who underwent a transplant respond toward abundant ROS by upregulating the transcription factor nuclear factor erythroid-derived 2-like 2, which regulates multiple stress-responsive pathways that protect from oxidative stress.²⁰ Administration of the enzymatic antioxidant thioredoxin 1in murine allo-SCT mismatch models (eg, GVHD) reduced T-cell (allo-) activation and, hence, GVHD without affecting GVL.²¹ However, the role of ROS-induced DNA damage on the donor T-cells’ ability for disease control remains unclear. In fact, a previous study has emphasized the importance of DNA damage response in T cells for proper immune surveillance in preclinical cancer models.²² To address this open issue, we investigated whether T cells from patients who underwent allo-SCT exhibit increased oxidative DNA damage and how this might affect their function and, consequently, clinical outcome.

Blood samples were retrieved from patients upon informed consent, in accordance with the Declaration of Helsinki (approval number of local ethic committee: 200_12, 280_14 B and 313_17B). Peripheral blood mononuclear cells were isolated using Ficoll-Paque (GE HealthCare). Patients’ characteristics are summarized in supplemental Table 1, available on the Blood website.

For flow cytometry (fluorescence-activated cell sorter [FACS]) analyses, samples were stained with fluorochrome-conjugated antibodies and chemical compounds. A complete list of antibodies is provided in supplemental Table 2. For immune-phenotypic characterization, fluorescent cell barcoding was carried out before extra- and intracellular antibody staining, as previously described in detail.²³ Samples were recorded on a FACS Canto II flow cytometer (BD Biosciences, Franklin Lakes, NJ) or on a Cytek NL-3000 full spectrum flow cytometer (Cytek Biosciences, Fremont, CA). Data were analyzed using FlowJo Version 10 (FlowJo LLC, Ashland, OR).

For more information, see the supplemental Methods.

First, we analyzed a total of 66 patients with hematologic malignancies who received an allo-SCT from HLA-matched (n = 58) or HLA-haploidentical donors (n = 8). The patients’ clinical characteristics are summarized in supplemental Table 1. Samples were collected at days +30, +45, +60, +75 +90, and +120 after allo-SCT.

Cells release oxidized cell-free DNA into the circulation, which is an indicator for oxidative stress.²⁴ We measured the serum concentration of 8-OHdG as a biomarker for oxidative DNA damage in healthy donors (HDs) (n = 18), patients who underwent autologous SCTs (auto-SCTs) (n = 16; median time point day +45 [29-46]), and patients who underwent allo-SCT (n = 50) using an enzyme-linked immunosorbent assay. Levels of circulating 8-OHdG were increased in patients who underwent allo-SCT at all time points, which is consistent with previous observations from the alloreactive environment,^18-20,25 whereas there was no difference between HDs and patients who underwent auto-SCT (Figure 1A). Next, we measured the 8-OHdG level in the T cells of HDs (n = 52), patients who underwent auto-SCT (n = 16; median time point day +43 [23-46]), and patients who underwent allo-SCT (n = 66) using FACS. A significantly increased 8-OHdG median fluorescence intensity (MFI) could be detected in those with allo-SCT T cells up to day +60 but not in patients who underwent auto-SCT (Figure 1B). This observation may suggest that ROS-triggering effects induced by high-dose chemotherapy are no longer effective at the time point of our sampling, and mechanisms associated with alloreactivity are more likely the cause of oxidative stress in patients who underwent allo-SCT beyond day +30.²⁵ Measurements of 8-OHdG levels in CD4⁺ and CD8⁺ T cells correlated positively (supplemental Figure 1A). In addition, we examined 8-OHdG levels in CD14⁺ monocytes and CD3⁻CD56⁺ NK cells. Here, only NK cells showed increased 8-OHdG accumulation, which correlated positively with that in T cells (supplemental Figure 1B-D). Typically, histone h2ax is phosphorylated in response to DNA damage.²⁶ Because 8-OHdG levels were significantly linked to ph2ax in allo-SCT T cells, our notion of increased DNA damage was further corroborated (Figure 1C). To exclude the possibility that our observations were affected by physiological fluctuations, we measured 8-OHdG and p2hax in T cells from HD (n = 12) once per week for 6 weeks without detecting significant time-dependent changes (supplemental Figure 1E-F). For subsequent analyses, we grouped patients who underwent allo-SCT based on their average 8-OHdG MFIs in T cells as high(hi), intermediate (int), and low (lo) groups (Figure 1D). Although 8-OHdG MFI T cells decreased continuously over time, the groups of patients with 8-OHdG^hi and 8-OHdG^lo levels differed significantly throughout the complete observational period (Figure 1E). In line with our findings suggesting oxidative stress after allo-SCT, protein carbonyls and thiobarbituric acid reactive substances (TBARSs) that result from oxidative protein and lipid damage are also significantly elevated in patients. However, we see no association of the serum 8-OHdG (Figure 1F), protein carbonyl, and TBARS concentrations with cellular 8-OHdG levels (supplemental Figure 1G-H), which may indicate that the extent of DNA damage in T cells depends not only on extrinsic ROS abundance but also on intracellular ROS production and cell-intrinsic protective mechanisms (eg, DNA damage repair). Regarding intracellular ROS production with consecutive DNA damage, alloreactivity of donor T cells after allo-SCT is certainly of great importance,²⁵ which we were able to model in in vitro cocultures of antigen-presenting monocyte-derived dendritic cells together with allogeneic T cells. T cells alloactivated in this setting showed a significant increase in 8-OHdG and phosphorylation of h2ax (supplemental Figure 1I). Moreover, our in vitro data do not indicate that extracellular 8-OHdG is substantially taken up by activated T cells (supplemental Figure 1J).

Based on their CCR7 and CD45RO expression, CD4⁺ and CD8⁺ CD3⁺ T cells were subdivided into naïve (CCR7⁺CD45RO⁻), central memory (CCR7⁺CD45RO⁺), effector memory (EM, CCR7⁻CD45RO⁺), and terminally differentiated effector (EMRA, CCR7⁻CD45RO⁻) cells (supplemental Figure 1K). Increasing evidence suggests that naïve T cells mediate GVHD, whereas memory T cells rather preserve beneficial T-cell functions, including the GVL effect.²⁷ Frequencies of naïve T cells did not differ between the 8-OHdG^hi and 8-OHdG^lo group. Although central memory CD4⁺ and CD8⁺ T cells had significantly increased in 8-OHdG^hi patients, CD4⁺ EM and CD8⁺ EMRA T cells had decreased (Figure 1G). Recovery of CD31⁺ recent thymic emigrant (RTE) CD4⁺ T cells that act as a surrogate for thymopoiesis was similar in both groups²⁸ (supplemental Figure 1L). Furthermore, T cells of both groups displayed an inverted CD4/CD8 ratio as compared with T cells of HDs, which is consistent with the more rapid CD8 reconstitution after allo-SCT (supplemental Figure 1M).²⁹ The frequency of naturally occurring CD25⁺CD127^neg CD4⁺ T_Regs, which have been previously shown to be more resilient toward ROS-induced cytotoxicity,¹² was higher in the 8-OHdG^hi group (Figure 1H).

Oxidative stress can affect hematopoietic stem cell (HSC) function.³⁰ Therefore, we evaluated the patients’ white blood cell engraftment, which was defined as the first of 3 consecutive days with a leucocyte count above 1000/μL as well as the course of the absolute neutrophil count, hemoglobin, and platelets. We could not detect any significant differences between the patient groups (supplemental Figure 1N-O).

Furthermore, we could not identify any demographic (eg, gender or age) or clinical (eg, remission status at transplantation or donor type) determinant for the 8-OHdG levels, as summarized in supplemental Table 3.

Next, we FACS-sorted 8-OHdG^hi (n = 4, from day +52 to +106) and 8-OHdG^lo (n = 5, from day +43 to +106) T cells, followed by RNA-sequencing analysis (GSE204782) (Figure 2A). Gene expression profiling together with the principal component analysis suggests a separation between both populations (Figure 2B-C; supplemental Table 4). Furthermore, our gene set enrichment analyses revealed that 8-OHdG^hi T cells, among other sets, are significantly enriched for genes that are involved in metabolism (eg, oxphos, glycolysis, and tricarboxylic acid cycle) and cell proliferation (eg, E2F targets, cell cycle, and G2M checkpoint) (Figure 2D). To investigate these phenomena in vitro, we treated T cells from HDs with low, sublethal ROS concentrations that still allowed adequate activation in terms of preserved proliferative capacity (supplemental Figure 2A-B). Consistent with the data from the ex vivo transcriptome analyses, treatment with ROS resulted in DNA damage but was simultaneously associated with greater activation of AKT, mTOR, and Myc signaling pathways (as assessed using FACS) that regulate cell division and metabolism (Figure 2E). In addition, we performed transcriptome analyses of HD-derived T cells treated with ROS and their untreated controls (GSE204781). Again, differences in gene expressions were observed, and we noticed an enrichment of genes associated with cell division and metabolic activity (supplemental Figure 2C-D). Finally, an increased expression of genes related to cell division in 8-OHdG^hi T cells isolated from patients who underwent allo-SCT was confirmed by a sustained higher frequency of proliferating Ki-67⁺ (Figure 2F-G). To investigate whether these differences between 8-OHdG^hi and 8-OHdG^lo patients also hold true within an individual patient who underwent allo-SCT, the patient's T cells were divided as the top 33% and the lowest 33%, based on their 8-OHdG MFI. Also in this analysis, cells with more 8-OHdG were shown to have a stronger Ki-67 expression than those with less 8-OHdG (supplemental Figure 2E). Furthermore, we observed in T cells activated in vitro that the 8-OHdG level is highest during the replicative S phase. In fact, T cells appear to accumulate 8-OHdG during each of their divisions, which may indicate a particular sensitivity of (allo-)activated, proliferating T cells toward ROS-triggered DNA damages (supplemental Figure 2F).

To follow up on the increased levels of metabolism-related transcripts in 8-OHdG^hi T cells (Figure 3A), we investigated additional metabolic markers using FACS. After allo-SCT, the glycolytic pacemaker enzyme hexokinase-2 was significantly higher expressed in 8-OHdG^hi than in 8-OHdG^lo T cells (Figure 3B). Fittingly, the uptake of the fluorescently labeled glucose analog 6-NBDG was also stronger, and we found higher levels of the proglycolytic transcription factor hypoxia inducible factor 1α (HIF-1α) in 8-OHdG^hi than in 8-OHdG^lo T cells (supplemental Figure 3A-B). Again, we found higher hexokinase-2 and HIF-1α levels in T cells with more 8-OHdG within 1 patient who underwent allo-SCT (supplemental Figure 3A-B). We could not detect significant differences in terms of carnitine-palmitoyl-transferase 1a expression, which controls the rate-limiting step of fatty acid oxidation or mitochondrial biogenesis, between both groups (supplemental Figure 3E-F). The mitochondrial membrane potential, which can inversely correlate with metabolic fitness,³¹ was lower in 8-OHdG^hi T cells (supplemental Figure 3G). Again, it was possible to induce in vitro a similar metabolically more active phenotype in T cells (ie, an enhanced uptake of glucose and fatty acids and an increased expression of the corresponding transporter) by treating using low-dose ROS (supplemental Figure 3H).

To further confirm our findings using an unsupervised approach, we analyzed our data with t-distributed stochastic neighbor embedding (t-SNE). Each event (ie, single cell) is assigned 2 new variables (ie, t-SNE 1 and t-SNE 2), and all fluorochromes are simultaneously classified as discriminants. Patient samples from the 8-OHdG^hi (n = 33) and 8-OHdG^lo groups (n = 33) as well as HD (n = 54) were analyzed at all time points using t-SNE. The events of the 3 groups (HD, 8-OHdG^hi, and 8-OHdG^lo) clustered in different areas, and when evaluating areas of high density (gates labeled with 1, 2, and 3), levels of carnitine-palmitoyl-transferase 1a and HK-2 expression were higher for 8-OHdG^hi T cells than for the other T cells (Figure 3C). Next, we used the SCENITH approach to functionally assess the T cells’ metabolic phenotype using FACS.³² Matching the t-SNE analyses, the 8-OHdG MFI correlated positively with protein synthesis (ie, puromycin incorporation), which has been previously shown to act as a surrogate for overall metabolic activity (supplemental Figure 3I). However, further analyses of the use of metabolic inhibitors did not show any metabolic skewing toward either glycolysis or OXPHOS in the 8-OHdG^hi group (supplemental Figure 3J).

Usually, proliferation and metabolic activity correlate with T-cell activation. Therefore, we evaluated the prototypical T-cell activation markers CD25, CD69, and CD137 on reconstituting 8-OHdG^hi and 8-OHdG^lo T cells at every time point. CD25, CD69, and CD137 levels were significantly higher in the 8-OHdG^hi group at every time point than in the other groups (Figure 4A-C; supplemental Figure 4A-C), further supporting the link between T-cell activation, proliferation, and metabolism. As already shown for other parameters, in individual patients who underwent allo-SCT, the relationship between the 8-OHdG level and the activation markers CD69 and CD137 was comparable (supplemental Figure 4D). Furthermore, patient samples from the 8-OHdG^hi (n = 33) and 8-OHdG^lo groups (n = 33) as well as samples from HDs (n = 54) were analyzed using t-SNE (for each time point individually). The events of the 3 groups (HD, 8-OHdG^hi, and 8-OHdG^lo) clustered in different areas, and when evaluating areas of high density (gates labeled with 1, 2, and 3), levels of CD69 and CD137 expression were higher for 8-OHdG^hi T cells than for other T cells (Figure 4D). Of note, the activation phenotypes of T cells in 8-OHdG^hi and 8-OHdG^lo patients share many similarities early after allo-HSC transplantations but become increasingly different over time (Figure 4C). In accordance with our findings, treating T cells during stimulation with low-dose ROS further boosted expression of the activation marker (Figure 4E).

Immune exhaustion can lead to an inadequate GVL reaction.⁵ Inhibitory signals promote T-cell exhaustion and can also be regulated by ROSs.³³ Therefore, we investigated the coinhibitory receptors PD-1, killer-cell lectin-like receptor G1 (KLRG1), CD152 (CTLA4), and CD223 (LAG3) on 8-OHdG^hi (n = 6-10) and 8-OHdG^lo T cells (n = 6-10). PD-1 and KLRG1 levels were significantly elevated in 8-OHdG^hi T cells, whereas we observed a similar trend for CD152 and CD223 (Figure 5A; supplemental Figure 5A). When treating HD-derived T cells with low-dose ROS during polyclonal activation, we found a comparable upregulation of immune checkpoint molecules (Figure 5B). Next, we assessed the 8-OHdG^hi (n = 6) and 8-OHdG^lo (n = 6) T cells’ response toward polyclonal stimulation, in terms of IFN-γ and granzyme-B (GrzB) production. The 8-OHdG^hi group showed significantly lower IFN-γ level and a tendency toward less GrzB expression (Figure 5C; supplemental Figure 5B).

To determine whether these dysfunctions are associated with deficits in eliminating malignant cells, patient-derived 8-OHdG^hi and 8-OHdG^lo cells (both n = 3) were cocultured with 2 AML cell lines (MOLM-13 and OCI-AML3), T cells were stimulated, and cell death of AML cells was assessed. The ability to clear AML cell lines was superior for 8-OHdG^lo T cells than for 8-OHdG^hi T cells (supplemental Figure 5C). To mimic the in vivo situation more accurately, in the next experiments, we treated the primary AML blasts isolated at the point of initial diagnosis with the corresponding reconstituting 8-OHdG^hi and 8-OHdG^lo cells (both n = 3) after allo-SCT. The killing of AML blasts was assessed using FACS and live cell imaging. Again, we observed a significantly reduced potency of 8-OHdG^hi T cells to eliminate (primary) leukemic targets (Figure 5D-E). We could confirm our observations using ROS-pretreated T cells, which also lost their ability to target AML cell lines and to produce IFN-γ and GrzB (Figure 5F; supplemental Figure 5D).

Because γδ T cells are considered to be important mediators of the GVL response,³⁴ we determined their frequency in patients who underwent 8-OHdG^hi and 8-OHdG^lo allo-SCT. We could not detect any significant difference. What we did observe, however, was that γδ T cells exhibited less oxidation-induced DNA damage and, thus, lower p2HAX activation as compared with the rest of T cells (supplemental Figure 5E). This could indicate a stronger resilience toward oxidative stress. Another T-cell–associated factor important for GVL and GVHD is the diversity of the T-cell receptor repertoire.³⁵ From our (quantitatively limited) RNA seq data, we could not find any differences between the 8-OHdG^hi and 8-OHdG^lo groups (supplemental Figure 5F). To better understand the impact of 8-OHdG on T-cell function, we sought out to identify potential systematic signatures of double-strand DNA damage in the RNA seq from CD3⁺ T cells sorted from patients who underwent 8-OHdG^hi and 8-OHdG^lo allo-SCTs. Because RNA seq generates sequencing reads only from actively expressed genome regions, we focused on elucidating potential gene fusion events and their putative underlying the chromosomal rearrangement as surrogate markers for double-strand DNA damage. We especially investigated fusions in selected genes relevant for T-cell function that would preferentially occur within the 8-OHdG^hi group. To do so, we applied the current version of the transcriptome-based fusion gene detection pipeline, JAFFA 2.0.³⁶ In fact, using JAFFA, we found 14 rearrangement genes relevant for T-cell function (ie, kinases: eg, ZAP70, FYN; effector molecules: eg, GZMA, GNLY, KLRG1; transcription factors: eg, NFATC2, RUNX3; and cytokine receptors: eg, IL7R, IL21R). From these rearrangement events, we generated a circos plot, showing all events and individual sites of rearrangement (supplemental Figure 5G). At this point, one can speculate whether those fusion events are directly responsible for the documented T-cell dysfunctions.

Taking into consideration our findings on the reduced in vitro functionality of 8-OHdG^hi T cells, we were interested in elucidating the impact of oxidative DNA damage in reconstituting T cells for our patients’ clinical courses. Firstly, we studied the 66 patients who underwent allo-SCT with a median follow-up time of 1027 days (supplemental Table 1). As described in “Increased levels of 8-OHdG in reconstituting T cells after allo-SCT,” patients were grouped based on the average 8-OHdG MFIs in T cells from all time points (Figure 1D). When comparing patients of the 8-OHdG^hi and 8-OHdG^lo group, we noticed that 8-OHdG^hi patients had an increased relapse incidence and a significantly shorter overall survival (OS) (Figure 6A-B). Furthermore, higher 8-OHdG levels were not associated with the occurrence of acute GVHD (of any grade), cytomegalovirus, and/or Epstein-Barr virus (re)infection (supplemental Figure 6A-C). To identify the most critical time point of increased DNA damage in T cells as a predictor for relapse and OS, we performed receiver operating characteristic (ROC) curve analyses. The highest area under the curve values for relapse and OS were achieved, for the 8-OHdG levels in T cells, at day 60 after allo-SCT (Figure 6C). Interestingly, day 60 marks the time point when 8-OHdG MFI in T cells normalize in the 8-OHdG^lo group but remains significantly elevated in the 8-OHdG^hi cohort (supplemental Figure 6D). Next, we expanded our cohort to a total of 100 patients with allo-SCT, with a median follow-up time of 887 days. The data of the additional patients are summarized in supplemental Table 5. Again, patients were divided into 3 groups based on the 8-OHdG MFI at day 60. As suggested by our ROC results, patients in the 8-OHdG^hi subgroup displayed a significantly higher cumulative relapse incidence and a shorter OS than those in the OHdG^lo subgroup, further corroborating our initial observations (Figure 6D-E). We could not detect any correlation between the 8-OHdG MFI in T cells at day 60 and the grade of acute GVHD (supplemental Figure 6E). After elaborating on the significance of day 60, we used an extended panel (of 18 markers) for full-spectrum flow cytometry to analyze samples that had been retrieved around day 60 from patients who underwent 8-OHdG^hi (n = 15) and 8-OHdG^lo allo-SCTs (n = 15). In the following t-SNE analyses, we found, consistent with our previous observations, that in clusters characteristic for a patient who underwent an 8-OHdG^hi allo-SCT, T cells show strong expression of both activation and metabolic markers, which are shown together with the T-cell composition (suppl. Figure 6F-G).

Disease relapse remains the major cause of mortality among patients who underwent allo-SCTs. Therefore, it is obligatory to better understand the underlying pathomechanisms to further improve therapeutic efficacy. In this context, a number of immune escape mechanisms have been described to limit GVL efficacy.⁵ They include the expression of immunological checkpoint molecules, metabolic perturbations, and premature immune cell exhaustion.^8,9,37 Here, we investigated whether oxidative stress is detrimental for the functional reconstitution of T cells and, consequently, their ability to mediate GVL. It is well established that oxidative stress, as a hallmark of cancer, contributes to tumor immune escape.¹⁴ The abundance of ROS negatively affects T-cell function and survival.^13,38 In fact, prevalence of oxidative stress has been previously reported in allo-SCT,^18-20,25 which we confirmed with increased 8-OHdG, protein carbonyls, and TBARS in the patients’ serum as well as 8-OHdG in circulating T cells.

Despite previous reports on the detrimental effects of oxidative stress on the function of HSCs,^30,39 we could not detect any association between the oxidative DNA damage in T cells and hematopoietic engraftment. In fact, resistance to (oxidative) stress is considered a key characteristic of stem cells.⁴⁰ At the same time, we observed higher frequencies of more mature T-cell subsets (ie, EM and EMRA T cells) in the 8-OHdG^lo patients, which is very well in line with a decreasing resilience toward ROS with increasing T-cell differentiation.⁴¹ In addition, faster engraftment and sustained higher levels of T_Regs, a cell type that is characterized by superior (as compared with conventional T cells) resistance toward ROS, were observed in patients who underwent 8-OHdG^hi vs those who underwent 8-OHdG^lo_allo-SCTs.^12,42 

Transcriptomic characterization of T cells with varying degrees of oxidative DNA damage revealed an enrichment of transcripts related to bioenergetic metabolism and cell cycle in the 8-OHdG^hi group, which we confirmed via analyses on the protein and functional levels. In fact, impairment of these pathways (in T cells) has been linked to AML relapse after allo-SCT.¹² In contrast, it has been shown that blocking Akt/mTOR signaling in T cells, which among other functions controls activation, proliferation, and metabolic activity, can improve immune therapies.^43,44 Similarly, inhibition of glycolytic metabolism that can either be promoted directly by Akt/mTOR or mTOR-controlled HIF1α- or Myc-activation enhances the T cells’ antitumor function.^45-47 This is in line with observations from adoptive T-cell therapy optimization studies that indicate that (mTOR-driven) glycolytic activity may favor terminal differentiation and senescence of T cells.⁴⁵ In this context, it is also important to note that optimal T-cell activation requires mitochondrial ROS production.¹⁷ Moreover, ROS can act as amplifiers of cell signaling by promoting phosphorylation of Akt,⁴⁸ which we could confirm when treating T cells with (low-dose) hydrogen peroxide, and it may explain why reconstituting 8-OHdG^hi T cells express markers associated with activation, proliferation, and metabolic activity more strongly than other T cells.

Chronic T-cell stimulation can lead to exhaustion.⁴⁹ A key feature of exhausted T cells is the sustained expression of inhibitory receptors, such as PD-1 or KLRG1.⁵⁰ This represents a compensatory mechanism to avoid immune overactivation. We found higher levels of both PD-1 and KLRG1 on 8-OHdG^hi T cells. As expected, the T cells’ ability to produce IFN-γ and eliminate AML blasts was significantly impaired. Of course, the latter findings might also have been partly caused by the T cells’ increased exposure to ROSs, as a result of ROS-induced cell damage and senescence.^12,38,41 At the end, both increased expression of inhibitory receptors and T-cell dysfunctions are associated with relapse after allo-SCT.^8,51-53 Indeed, we observed higher relapse rates and a shorter OS in the 8-OHdG^hi patient group. Incidence of GVHD was not increased, which would tend to be against the narrative that alloreactivity of T cells, which leads to increased mitochondrial ROS production,⁵⁴ is causative for the 8-OHdG^hi T-cell phenotype. In addition, no increased rate of cytomegalovirus or Epstein-Barr virus reactivation could be detected, indicating a preserved graft-versus-infection effect. Our ROC analyses identified 8-OHdG^hi levels at day +60 after allo-SCT to be the most suitable time point to predict relapse and OS, with the average time to relapse being +282 days.

The mechanisms leading to oxidative stress after allo-SCT remain to be elucidated. Genetic polymorphisms in the oxidative stress pathways of the host and/or the donor could determine the amount of ROS production and/or resilience toward ROS-mediated (DNA) damage. In fact, 8-OHdG glycosylase (OGG1) is primarily responsible for the removal of 8-OHdG from damaged DNA bases once it has been formed.⁵⁵ Polymorphisms at position 1245 in exon 7 of the human OGG1 lead to a substitution of cysteine for serine and are associated with a lower DNA repair activity,⁵⁶ which also could favor 8-OHdG accumulation and need to be investigated, especially in donor cells. Theoretically, we could then select third-party stem cell donors based on the (most favorable) OGG1 polymorphism, but obviously this requires further research on sufficiently large donor cohorts together with the corresponding allo-SCT outcome data. Recently, a first-of-its class pharmacological activator of OGG1 was introduced.⁵⁷ This could allow us a targeted repair of 8-OHdG, which could take place either by the systematic administration of an OGG1 activator or in the context of pretreating donor lymphocyte infusions. Most likely, both oxidative stress level and cell intrinsic protective mechanisms codetermine oxidative DNA damage in T cells, because we could not detect a direct correlation between cellular oxidized DNA contents and circulating oxidized nucleic acids, lipids, and proteins as surrogates for systematic ROS abundance. Nevertheless, the use of systematic antioxidants such as thioredoxin-1 could be beneficial for patients who underwent allo-SCT, because preclinical studies describe an improvement of the immunometabolic T-cell phenotype in the context of antitumor responses, and at the same time, benefits in terms of GVHD development.^21,58 

In summary, we demonstrate that oxidative stress is associated with DNA damage in reconstituting T cells after allo-SCT. T cells with high levels of DNA damage showed signs of sustained activation and premature exhaustion. This T-cell phenotype was associated with functional deficits, a higher relapse incidence, and a shorter OS. Remarkably, T-cell DNA damages appear to specifically affect GVL activity but not GVHD. A better understanding of the underlying mechanisms (eg, genetic polymorphisms in oxidative stress pathways) could be relevant for donor selection or could lead to the development of new therapies aimed at improving DNA damage repair in T cells or at redox remodeling after allo-SCT.

The authors thank the patients, their families, and referring physicians who made this study possible. The authors thank Florentine Schonath (Core Unit Cell Sorting and Immunomonitoring Erlangen), Alina Kämpf, Lina Meretuk, Luisa Albert, and Dorothea Gebhardt for excellent technical assistance. The graphical abstract was created with BioRender.com.

F.K., C.F., R.J., H.B., E.H., T.D., A.M., and D.M. were funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation); Project-ID 324392634 - TRR 221. S.V and A.M. were supported by the Deutsche José Carreras Leukämie-Stiftung (02 R/2021).

Contribution: F.K., C. Liang, and R.B.-L. planned and performed research, compiled and analyzed data, and helped writing the manuscript; A.S., C.F., C. Lischer, S.R., S.V., B.J., M.B., R.J., W.R., H.B., T.D., and A.M. performed research, analyzed data, and provided material; T.F. and E.H. helped writing the manuscript; and D.M. designed the study, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: D. Mougiakakos, Department for Hematology and Oncology, Otto-von-Guericke University, Leipzigstr 44, 39120 Magdeburg, Germany; e-mail: dimitrios.mougiakakos@med.ovgu.de.